Office Hours

[agcohn821]

[9:00 AM] beta_helix: Hi everyone! Welcome to a Cryo-EM Density Foldit Office Hours!
[9:01 AM] Dudit: Hi @beta_helix
[9:01 AM] Susume: hi beta
[9:01 AM] beta_helix: Hi Dudit: and SusumeL
[9:01 AM] Huubr: Hi @beta_helix:. Glad I could make it, just in time :-)
[9:02 AM] beta_helix: Perfect timing! So I wanted to have a specific chat about Puzzle 1964, since it's different than anything we've ever posted before:
[9:02 AM] beta_helix: https://fold.it/portal/node/2011305
[9:03 AM] Dudit: @beta_helix: Are there a plan in Foldit to add Selenocysteine Amino Acid ?
[9:03 AM] Dudit: https://en.wikipedia.org/wiki/Selenocysteine
Selenocysteine (symbol Sec or U, in older publications also as Se-Cys) is the 21st proteinogenic amino acid. Selenocysteine exists naturally in all three domains of life, but not in every lineage, as a building block of selenoproteins. Selenocysteine is a cysteine analogue with a selenium-containing selenol group in place of the sulfur-containin…
[9:03 AM] Susume: I am loving 1964, no idea how others are finding it
[9:04 AM] beta_helix: Good question, Dudit:. We've discussed adding non-canonical amino acids before, but never pursued that idea.
[9:04 AM] beta_helix: I'm happy to hear that, Susume
[9:05 AM] Huubr: I find 1964 very interesting! I am curious how an ED map is created. Could you tell us something about that?
[9:05 AM] beta_helix: As background, we had an exciting Foldit publication demonstrating how good you all are at fitting cryo-EM models: https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.3000472
Building de novo cryo-electron microscopy structures collaborativel…
This Community Page article demonstrates that microscopists can now collaborate with the players of the computer game Foldit to generate thorough high-quality de novo structural models; this development could greatly speed the generation of excellent Cryo-EM structures when used as a complement to current methods.
[9:07 AM] beta_helix: @HuubR That is a great question. I can briefly explain (as I am not an experimentalist) but I think with cryo-EM, it's a lot cooler (pun intended!) to see how it is created. IMO, this short 1 minute YT video is a pretty good explanation: https://www.youtube.com/watch?v=6G550DfY75Q
YouTube
MRC Laboratory of Molecular Biology
Cryo-EM Animation
[9:09 AM] beta_helix: The advantage of cryo-EM over x-ray crystallography is that crystalizing proteins means you lose any movement. Cryo-EM is able to freeze a protein using liquid ethane, without creating any ice crystals. Then a bunch of 2D images are taken and reconstructed into a 3D model with computers
[9:10 AM] beta_helix: Previously in Foldit we were always able to give you a small Electron Density (ED) cloud, because we knew the answer and could easily trim the density for you.
[9:11 AM] beta_helix: (That was the case for the 4 puzzles from that paper) But this is the first puzzle where we can't provide you with that information!
[9:11 AM] Huubr: A bunch of 2D images converted to a 3D model, as in "regular" X-ray tomography?
[9:12 AM] Susume: you did once give us a big brick of cloud in which we had to choose a copy of the protein, but it was still a monomer and they were all lined up the same way
[9:12 AM] beta_helix: So the 2D images are in many different orientations, thankfully, making it easier to build a 3D model (rather than if they were all oriented the same way).(edited)
[9:13 AM] beta_helix: Yes, Susume good point. The purpose of that puzzle was to see if the tools in Foldit were sufficient, or if we needed to provide you with a trimmed map.
[9:13 AM] beta_helix: If I recall, it was thanks to that puzzle that the Trim Tool was requested and added to the game
[9:15 AM] Susume: if we get a good model of most of the protein but there are a couple of loops for which we can't match the density, is it possible to solve (or partially solve) the protein from that, or only if we get the whole thing right?
[9:16 AM] beta_helix: Great question!
[9:17 AM] beta_helix: If that is the case, we should be able to trim the density based on that "pretty darn close" model… and then we could repost it as a much easier ED puzzle.(edited)
[9:17 AM] beta_helix: At that point, we might even be able to include the RNA molecule!
[9:17 AM] Susume: I ask because the loop in the center of the donut is very hard to find a path for, as are the two tails
[9:18 AM] beta_helix: This puzzle is large enough already, that we hesitated about adding in more molecules (and slowing the puzzle down even more).
[9:19 AM] beta_helix: In that case, the RNA might help with that
[9:20 AM] Huubr: The density map that we have now has a regular pattern of (almost?) identical monomers in it. Would you know whether that periodicity can be used to enhance the resolution? In other words, can the repetition in the pattern be exploited to reduce the noise, by averaging it out? Or is that something that has already been done?
[9:21 AM] beta_helix: @HuubR Wow, that is a great question that I do not know that answer to. My assumption is that our collaborators gave us the very best map they currently have.
[9:21 AM] Susume: maybe - but I think I found the RNA, it looks to be buried fairly far from the inner donut surface, and it is the very surface I am struggling with
[9:21 AM] beta_helix: Perhaps by providing a decent model, however, they could somehow enhance the resolution (but I have no idea if that is possible).
[9:22 AM] beta_helix: @Susume I see… In that case my suggestion would be to follow the score function, and see where Rosetta believes it fits best
[9:23 AM] Susume: the repetition can be used by players - if you find a specific piece that is repeated periodically, you must fill exactly one copy of that piece with protein
[9:23 AM] beta_helix: I will be sending today's transcript to our collaborators and will let you know their feedback!
[9:23 AM] Susume: (unless you think it is the RNA in which case don't fill it)
[9:24 AM] beta_helix: Question for you all: have you been able to trim down the density to a single monomeric area (-ish)
[9:24 AM] beta_helix: to make it easier to work with?
[9:25 AM] Huubr: You mean, trim in our heads, or on the screen?
[9:25 AM] beta_helix: on the screen
[9:25 AM] Susume: in my head yes - I forgot to try using the actual trim tool ….
[9:25 AM] beta_helix: I was referring to the Radius Slider: https://fold.it/portal/node/2009618
New Release!
New Release!
[9:26 AM] beta_helix: * Added "radius" slider to electron density panel to view only electron density data within a ball of the given radius.
[9:26 AM] beta_helix: I feel like this puzzle would be so difficult without using that tool!
[9:27 AM] Susume: I tried using the sphere but then you can't see the repeats (or in my case the red dots I marked them with) to use as don't-go-here' landmarks
[9:27 AM] Huubr: I had found the Radius slider, but I would rather have an eraser :P
[9:27 AM] beta_helix: Thank you both, that is very helpful.
[9:27 AM] beta_helix: The eraser has been requested for a long time, but is not a trivial tool to implement.
[9:28 AM] Huubr: I had figured it would be difficult to implement a GUI for it.
[9:28 AM] Susume: the trim slider may be more useful - I'll work with it some. It still has the flaw of not saving your setting and not giving a number to your setting
[9:29 AM] beta_helix: @HuubR it would almost have to be its own separate component
[9:29 AM] beta_helix: @Susume Oh! That is very useful to know and shouldn't be too difficult to implement (famous last words from someone who isn't a dev can you not quicksave your current state?)(edited)
[9:31 AM] Susume: quicksaving and restoring does not restore the could trim that was in effect when you quicksaved
[9:31 AM] Susume: *cloud
[9:32 AM] beta_helix: That isn't good at all!
[9:32 AM] beta_helix: We'll see if we can fix that soon.
[9:32 AM] jmbrownlee333:In the early days and in revisiting puzzles we are given 'easier' puzzles with known answers. These have training value. Could puzzles that are easier of this type be an option? So we can learn on the 'bunny hill'
[9:33 AM] Susume: I take one thing back - reducing the sphere leaves the dots visible, just hides the cloud they are assoc with
[9:33 AM] beta_helix: @jmbrownlee333 That was our motivation for https://fold.it/portal/node/2011219
[9:34 AM] beta_helix: @ Susume thanks!
[9:34 AM] Susume: I would not mess with attaching a trim value to a quicksave - just make it like the other sliders that show their current state when you re-open them and display a number for their current setting
[9:34 AM] beta_helix: @Susume got it, that makes sense.
[9:35 AM] Huubr: Puzzle 1847 was a nice level to start with (i had been folding for 2 months then :-)
[9:35 AM] beta_helix: @jmbrownlee333 or were you referring to Revisited puzzles in general? (not specifically ED)
[9:35 AM] Huubr: (and 1847 had a very detailed map!)
[9:36 AM] beta_helix: @HuubR Ah yes, that was an awesome puzzle: https://fold.it/portal/node/2009799
[9:37 AM] jmbrownlee333:Just referring to the 'bunny hill' idea in general. I guess I meant Ed puzzles in specific.
[9:37 AM] beta_helix: We need more Foldit player designs to solve first
[9:37 AM] jmbrownlee333:But solve them with cryoEM and give us a swath of density
[9:38 AM] beta_helix: @jmbrownlee333 I completely understand. I was just worried you meant that we don't have enough Revisited puzzles up in general. But, indeed, other than the Beginner ED puzzle, and that recent last one, we rarely give you a known structure. This is easily something we can fix!(edited)
[9:39 AM] jmbrownlee333:The Beginner ED puzzle is a perfect example. Thats great training for ED puzzles in general.
[9:39 AM] beta_helix: Great, thanks!
[9:40 AM] beta_helix: Have you heard any feedback from your teammates or other players about the difficulty of this puzzle?
[9:42 AM] Susume: I have not, but I'm not as plugged in as I used to be

[9:42 AM] argyrw:hi
[9:42 AM] argyrw::)
[9:42 AM] beta_helix: Hi argyrw
[9:42 AM] argyrw: you can I learn about the bonus points all time loose bonus of that
[9:43 AM] argyrw: how can I learn
[9:43 AM] argyrw: is like blind the game
[9:43 AM] argyrw:IMAGE: http://fold.it/portal/files/chatimg/irc_936722_1615481002.png

[9:43 AM] argyrw:IMAGE: http://fold.it/portal/files/chatimg/irc_936722_1615481006.png

[9:43 AM] argyrw:this is one example
[9:43 AM] beta_helix: Protein design makes folding that much more difficult.
[9:44 AM] argyrw:yes but the bonus points give all points in the game
[9:44 AM] argyrw:like the puzzle which I play
[9:44 AM] argyrw:IMAGE: http://fold.it/portal/files/chatimg/irc_936722_1615481079.png

[9:44 AM] beta_helix: You aren't just folding and scoring the protein based on the Rosetta/Foldit score, you also need to make sure that the metrics are satisfied or else your design will simply not fold.
[9:44 AM] argyrw:hydrogen bonds network what is that?
[9:45 AM] beta_helix: This post should help you: https://fold.it/portal/node/2010984#comment-43985
[9:45 AM] argyrw:wait to try copy paste thank you
[9:46 AM] beta_helix: It addresses all the metrics, and hbonds, in one post!
[9:47 AM] beta_helix: You can also just go to the Forum and it's the top stickied post
[9:47 AM] beta_helix: scroll down to Part 6
[9:48 AM] beta_helix: Any other questions for this Foldit Office Hour? They don't have to be about ED
[9:49 AM] Susume: baker lab just had a paper about pairwise decomposable BUNS penalty - have the foldit team looked at it?
[9:50 AM] spvincent: I was wondering about Buns at the ends of helices: are they artefacts?
[9:50 AM] Susume: (I will bug bkoep too - just excited to see a filter that could possibly get incorporated in the regular score)
[9:51 AM] beta_helix: Great question, I didn't know about it but I'm sure bkoep and neil (who are both in the Baker Lab) know all about it. I will ask them about it.
[9:51 AM] Huubr: @Susume, could you briefly explain "pairwise decomposable BUNS penalty" to a regular Foldit player?
[9:52 AM] beta_helix: @Susume my guess is that it's probably in a very recent version of Rosetta, so it will require Foldit merging with the latest code. That is non-trivial, but is on the horizon with the new small-molecule design features that are coming.
[9:52 AM] Susume: BUNS in foldit have to be a filter (not part of regular score function) because it depends on groups of AAs, not just on pairs of AAs (everything in regular score has to be a measure of one AA or of a pair)
[9:53 AM] beta_helix: Oh dear, I knew as soon as I said "non ED stuff is ok to ask" I would be flooded with design questions
[9:53 AM] beta_helix: @spvincent I don't want to give you the wrong answer, so I will ask the team today about this
[9:53 AM] Susume: a baker lab scientist figured out a way to calculate BUNS penalty and assign it to pairs of AAs - which might make it possible to put into regular score - then it will play better with scripts, be more continuous, etc
[9:54 AM] Huubr: Thanks, @SusumeL, I am getting the picture (I think).
[9:54 AM] beta_helix: argyrw will be happy about that as well, Susume!
[9:55 AM] spvincent: tx
[9:55 AM] jmbrownlee333:non science question. Will we ever get the promisec new website? Do ya know anything about that perchance?
[9:55 AM] Susume: best thing about putting it in score is then wiggle tool can 'see' it and optimize it at the same time as other score parts - wiggle does the cost/benefit work instead of us
[9:55 AM] jmbrownlee333: promised
[9:55 AM] Josh: jmb I can answer this one – we are actively developing and testing it
[9:55 AM] jmbrownlee333:ETA?
[9:56 AM] Josh: The delays are due to a bottleneck of developer time – UW keeps stealing our dev :P
[9:56 AM] jmbrownlee333:I see
[9:56 AM] Josh:Unfortunately, I don't have an ETA :/ all I can say is soon
[9:57 AM] beta_helix: I would guess that at the end of this month, start of next month, you should see a "devprev" version of it (just my guess)
[9:57 AM] beta_helix: @ Susume yes indeed, we always want to include as much as possible in the actual score function!
[9:58 AM] beta_helix: any last questions for Josh or I?
[9:58 AM] Susume: oo, in honor of Formula, anything you wish players would do more of, or do less of?
[9:59 AM] beta_helix: I usually give the same answers, for design: try to submit as many solutions that meet all the metrics as puzzle, etc.
[9:59 AM] beta_helix: But today I'll say: tell your friends about Foldit.
[10:00 AM] Huubr: Got no more questions right now. I will try and collect some thoughts about image enhancement, and put them in a PM, if that's OK?
[10:00 AM] beta_helix: @HuubR of course!
[10:00 AM] beta_helix: This might seem silly for the many of you that have been around for so long… but I think Josh: has done some amazing work in Foldit
[10:01 AM] beta_helix: Making the game a lot more accessible to new players. So, hopefully when the new website is settled, it'll be easier to keep interested new players. Just my 2 cents
[10:01 AM] Josh: @ Susume my request, as usual, is for anyone who's been playing more than 6 months to find a way to share their knowledge – on the wiki, a guide on the forums, anything
[10:01 AM] Susume: yes he has!! so grateful for Josh!
[10:01 AM] Josh: <3 !!
[10:01 AM] beta_helix: Just look at the Forum post I was able to give argyrw earlier!
[10:02 AM] beta_helix: Normally I'd frantically be searching for the news posts or wiki pages!!!
[10:02 AM] Huubr: https://fold.it/portal/node/2010984#comment-43985
[10:02 AM] jmbrownlee333:I second SusumeL's thought
[10:02 AM] Josh: That How to Foldit guide was 92 pages long in a google doc…
[10:02 AM] Josh: glad it's helpful :)
[10:02 AM] Huubr: Me too (or would that make three? ;-)
[10:03 AM] beta_helix: On that note (Josh: rules!) I'll end with: Thank you all for chatting today, thanks for playing that giant ED puzzle, and thanks for the great folding that you do! Keep it up!

[agcohn821]

[11:01 AM] neilpg628: Hello All, its @neilpg628 , ready to answer some questions!!
[11:01 AM] Dudit:Hello @neilpg628
[11:02 AM] pc: hi
[11:02 AM] cjddig: Hello, it's my first time to attend the Office Hour.
[11:02 AM] pc: @neilpg628 : In the 1968 puzzle, BUNS never appeared on the player's protein, only on the virus Amino Acids. was it a bug or was it a voluntary setting ?
[11:03 AM] neilpg628: Do you mean 1966?
[11:04 AM] pc: 1968 "Influenza HA Binder Design Competition"
[11:05 AM] neilpg628: That could have been an issue. but I only deal with the rigid linker puzzles so I'm not sure
[11:05 AM] Dudit: @neilpg628 Why Foldit is always have a lag/latency (especially when there is a lot of Objectives) ?
[11:05 AM] pc: ok thanks ^^ @Dudit
@neilpg628 Why Foldit is always have a lag/latency (especially when there is a lot of Objectives) ?
[11:06 AM] neilpg628: The scientific backend of Foldit is 3 million lines of Rosetta C++ that is installed as a bunch of binaries on your machine The code is not optimized for gameplay and is meant to run on large clusters or distributed computing projects like Rosetta@home Thus it can be quite slow to do a lot of the expensive computations necessary on your machine alone
[11:07 AM] Dudit: Thanks
[11:08 AM] cjddig: I often experience a lot of lags in the ED puzzle. Does ED cloud causes lags or is the performance of my computer not good? Is there a way to reduce the lags?
[11:09 AM] neilpg628: It is probably the ED cloud We are working on ways to make the computation of coordinates less intensive I don't think computer performance is relevant since we are trying to make a game for a huge range of platforms and configurations
[11:09 AM] pc: @neilpg628 When I mutate a starting protein (like a tripple helix), there is often Treonine that appear on core. I think it is maybe because some helixes are too close sometimes, but it happen often. Should I keep them, or why "mutate" choose them instead of a "good" hydrophobic sidechain ?
neilpg628: It is probably the ED cloud We are working on ways to make the computation of coordinates less intensive I don't think computer performance is relevant since we are trying to make a game for a huge range of platforms and configurations
[11:09 AM] cjddig: OK, Thanks
[11:11 AM] pc: (so I fix thoses threonines with hand folding + wiggle often)
[11:14 AM] pc: oh he is disconected I think
[11:14 AM] donuts554: hello
[11:16 AM] HuubR: @pc, which residues do you have in your helices before mutate? When you start with Alanine, you are likely to get your helices very close together.
[11:17 AM] alcor29: There appears to be a UWIT outage.
[11:18 AM] neilpg628: How do you mean?
[11:18 AM] alcor29: oh you're bak.
[11:18 AM] neilpg628: I never left lol
[11:18 AM] neilpg628: I was just waiting
[11:20 AM] nspc: IMAGE: http://fold.it/portal/files/chatimg/irc_959223_1616865644.png

[11:20 AM] nspc: sometilmes I just start like this
[11:21 AM] cjddig: Another question: What is the criterion of the invalid residues? Alanine is good to make the core, but many puzzles consider it as invalid residue.
[11:21 AM] nspc: IMAGE: http://fold.it/portal/files/chatimg/irc_959223_1616865717.png

[11:22 AM] Dudit: @neilpg628 Is there a plan from the Foldit Developer Team to further optimizing Foldit to greatly reduce lag?
cjddig: Another question: What is the criterion of the invalid residues? Alanine is good to make the core, but many puzzles consider it as invalid residue.
[11:24 AM] neilpg628: Rosetta intrinsically favors some residue types over others ALA is one of these and if it was unrestricted then the most energetically favorable configuration would just be a bunch of ALA which would not be good
Dudit: @neilpg628 Is there a plan from the Foldit Developer Team to further optimizing Foldit to greatly reduce lag?
[11:24 AM] neilpg628: There are some optimizations that can be done but a lot of it is on the Rosetta side which is harder to change
neilpg628: Rosetta intrinsically favors some residue types over others ALA is one of these and if it was unrestricted then the most energetically favorable configuration would just be a bunch of ALA which would not be good
[11:25 AM] cjddig: Thanks
[11:25 AM] donuts554: How strong are the hydrogen bonds involving sulfur compared to pi stacking?
[11:26 AM] neilpg628: Pi stacking is probably weaker since Hbonds form the majority of electrostatic interactions
[11:27 AM] pc: To create a first stable protein alone in a design puzzle (or linker), I use the Building Blocs, Ideal SS… I worked on the core first (selecting only core) with mutate core only, and global wiggles. It seems to me that we should not mutate sidechain outside the protein at begin, because the bouds which are made could distort the protein too much when I wiggle. So I mutate outside only at end, when core it stable. Do you think that's ok, or is it better to mutate and wiggle all the protein each time ?
[11:29 AM] donuts554: But in the Foldit puzzles I dont see hydrogen bonds involving sulfur
[11:30 AM] donuts554: Why arent hydrogen bonds involving sulfur shown in the Foldit puzzles?
donuts554: Why arent hydrogen bonds involving sulfur shown in the Foldit puzzles?
[11:30 AM] neilpg628: Did you mean disulfides?
[11:30 AM] neilpg628: Those are chemical bonds and would be stronger than any Hbond or Pi stacking
[11:31 AM] donuts554: Not just disulfides
[11:31 AM] cjddig: Is sulfur polar atom which can make HBonds? I assume he is saying this.
[11:31 AM] neilpg628: I don't think sulfur can make Hbonds Not sure though
[11:32 AM] donuts554: Oh ok
[11:32 AM] donuts554: I think you forgot to answer Dudit:'s and pc's questions
[11:32 AM] jmbrownlee333: I think S-H covalent bonds engage in very weak hydrogen bonds and can thus be disregarded.
pc: To create a first stable protein alone in a design puzzle (or linker), I use the Building Blocs, Ideal SS… I worked on the core first (selecting only core) with mutate core only, and global wiggles. It seems to me that we should not mutate sidechain outside the protein at begin, because the bouds which are made could distort the protein too much when I wiggle. So I mutate outside only at end, when core it stable. Do you think that's ok, or is it better to mutate and wiggle all the protein each time ?
[11:34 AM] neilpg628: I think your strategy is fine Too much mutation can prevent any real structure from forming
[11:34 AM] pc: thanks
[11:35 AM] donuts554: But how weak are the weak hydrogen bonds that S-H covalent bonds engage in compared to pi stacking and the "Cation-Pi Interaction"?
[11:36 AM] neilpg628: I don't know the specific bond energies but they might be similar @bkoep might know for sure
[11:37 AM] pc: @neilpg628 In a design puzzle, player protein need hydrophibics that touch the virus (to stay in place). But if there is too much hydrophobics outside the player protein, it can have problems to fold alone. How many hydrophobic can we have outside, is there a metric for this that will be done ? Or it is a too much complex prediction to do (and maybe we will need "alpha fold" ?) If there are spaced enough do you think it is better ?
[11:38 AM] pc: (sorry for my long questions ^^ )
[11:39 AM] donuts554: But bkoep said that he didnt know that answer at the top of his head at a previous office hour
[11:40 AM] neilpg628: I'm not a biochemistry student so you might have to look around ¯_(ツ)_/¯
[11:41 AM] donuts554: Oh ok then
pc: @neilpg628 In a design puzzle, player protein need hydrophibics that touch the virus (to stay in place). But if there is too much hydrophobics outside the player protein, it can have problems to fold alone. How many hydrophobic can we have outside, is there a metric for this that will be done ? Or it is a too much complex prediction to do (and maybe we will need "alpha fold" ?) If there are spaced enough do you think it is better ?
[11:43 AM] neilpg628: I agree that too my hydrophobics is bad for folding but ideally the surface is mostly hydrophobic with a few polars that can hbond with polars on the binders for specificity The ContactMS and Core Exists Objectives try to capture some of this behavior.
[11:44 AM] Dudit: @neilpg628 In puzzle 1968 (Binder Design Competition), where are the final rankings?
[11:44 AM] pc: thanks. One of my strategy is to use hydrophobics that protrude from the core as possible
Dudit: @neilpg628 In puzzle 1968 (Binder Design Competition), where are the final rankings?
[11:45 AM] neilpg628: Are they not on the website?
neilpg628: Are they not on the website?
[11:46 AM] Dudit: I mean the total valid submission to the Scientists
[11:47 AM] neilpg628: I'm not sure what you mean but the puzzle only closed yesterday so they might still be being processed
[11:48 AM] donuts554: When will colored text in the Recipe Output dialog box be added to Foldit?
donuts554: When will colored text in the Recipe Output dialog box be added to Foldit?
[11:48 AM] neilpg628: @josh might know
[11:49 AM] alcor29: What are you currently working on neil?
[11:50 AM] neilpg628: My focus is the BUNS and DLP stuff(edited)
[11:50 AM] donuts554: Oh ok
[11:50 AM] alcor29: DLP?
[11:50 AM] neilpg628: The Designable Linker stuff
[11:51 AM] pc: @neilpg628 : In linker puzzles one of he main difficulty is to join the last cutted point (and keep ideality) do you plan to do a new tool to help for it ? there is some strategies (using band on helix), but it is still difficult.(edited)
[11:51 AM] donuts554: Oh wait I found a problem with the BUNS objective
[11:51 AM] donuts554: I remember that it was on a symmetry puzzle
[11:52 AM] donuts554: Yes it was on Puzzle 1954
pc: @neilpg628 : In linker puzzles one of he main difficulty is to join the last cutted point (and keep ideality) do you plan to do a new tool to help for it ? there is some strategies (using band on helix), but it is still difficult.(edited)
[11:53 AM] neilpg628: What sort of tool would be helpful?
[11:53 AM] donuts554: The BUNS objective in 1954 is not consistent, as 4 BUNS suddenly pop up from 0 BUNS when I rotate my green protein, and my green protein is far away from the other three locked proteins
[11:54 AM] pc: maybe when we try to join a cutted point, have a different representation of the 2 parts. To help us to know it is will be ok.
[11:54 AM] donuts554: And I rotated by left-clicking on the protein, and by dragging the purple arrow about 7 degrees clockwise
[11:55 AM] nspc: IMAGE: http://fold.it/portal/files/chatimg/irc_959223_1616867704.png
[11:55 AM] nspc: IMAGE: http://fold.it/portal/files/chatimg/irc_959223_1616867736.png
[11:56 AM] nspc: it is often bad when we join the last cutted point like this
[11:56 AM] nspc:(in linker puzzles)
[11:56 AM] Susume: I think donuts may be running into the voxel issue
donuts554: The BUNS objective in 1954 is not consistent, as 4 BUNS suddenly pop up from 0 BUNS when I rotate my green protein, and my green protein is far away from the other three locked proteins
[11:56 AM] neilpg628: This is a known issue unfortunately Basically the calculator used to compute surface area is sensitive to rotation of the structure even with no change This means that atoms can be flagged as 'buried' or 'not buried' depending on their orientation which affects the BUNS count
[11:58 AM] cjddig: How BUNS affect real protein? Why are they bad for protein?
[11:59 AM] Josh: donuts554 colored recipe text is currently a low priority because not many players are saying it would really benefit their play. If more players tell me they really want this feature, I can make it a higher priority.
[11:59 AM] Josh: *recipe output text
[12:00 PM] donuts554: Ok so thats why I dont want to play puzzles with the BUNS objective in them as much
[12:00 PM] donuts554: Oh ok Josh
[12:00 PM] Susume: pc I have a suggestion: make a new cut in the middle of a helix, then heal the cut in the loop, and use ideality or blueprint to make the loop healthy. Then you can work on healing the cut in the helix using bands - it is much easier in the middle of the helix to see if the shape will be good
cjddig: How BUNS affect real protein? Why are they bad for protein?
[12:00 PM] neilpg628: The blog post here might help They can prevent the protein from folding because these atoms would rather bind with water than remain buried https://fold.it/portal/node/2009053
The BUNS Objective
The BUNS Objective
[12:00 PM] donuts554: Also Neil what things do you want to encourage Foldit players to do, and what things do you want to discourage Foldit players to do?
neilpg628: The blog post here might help They can prevent the protein from folding because these atoms would rather bind with water than remain buried https://fold.it/portal/node/2009053
[12:00 PM] cjddig: Thanks, I'll look into it.
[12:01 PM] Skippysk8sIRC: In last puzzle with buns, I couldn't easily mutate the "mutatable" side chain on the frozen SS. I ended up using Maaaa on one segment, which seemed very slow
[12:01 PM] Skippysk8sIRC: Any better ideas?
[12:02 PM] alcor29: Linking LCB seems to intiuitively cut down the chances of bonding to the targer. Is the precise angle we use derived from the distribution on targets on the virus sphere? And are these targets symmetrically distributed like the tetrahedral bonds of a Carbon atom?
[12:02 PM] pc: thanks susume. I already tryed, it is not very easy but can works yes. I try to think about a tool that make the puzzle easier for the most people too ^^.
alcor29: Linking LCB seems to intiuitively cut down the chances of bonding to the targer. Is the precise angle we use derived from the distribution on targets on the virus sphere? And are these targets symmetrically distributed like the tetrahedral bonds of a Carbon atom?
[12:07 PM] neilpg628: The angle is pretty accurate and we took it from a very precise crystal structure of the actual binding complex
[12:07 PM] neilpg628: Thanks for all the questions!

[agcohn821]

1:59 PM] jflat06: Hello
[1:59 PM] pc: @jflat06 first question : how are the new features that are made for foldit decided? scientific needs, players sugestions, intern testing, other things ?
[2:00 PM] jflat06: @pc: pretty much a mix of those things! But our priorities usually place a premium on scientific needs
[2:00 PM] jflat06: usually shortly followed by issues involving the integrity of the competition
[2:01 PM] pc: thanks
[2:01 PM] jflat06: Then we like to tackle high-value features, whether they be player suggestions or otherwise
[2:02 PM] jflat06: One of the things that makes foldit pretty interesting/unique is that we have 4-5 different labs around the country all collaborating on the project, and each them has their own priorities and vision of how to use foldit
[2:03 PM] donuts554: hello
[2:03 PM] jflat06: hi
[2:03 PM] cjddig: When does new Foldit website come out? What is advanced or new features it has?
[2:03 PM] donuts554: Hello how are you?
[2:04 PM] jflat06: The website should hopefully be coming out within the next month, at least to a test site that players can play with. We're at the stage where we're reviewing the code to make sure its all good, and fixing/implementing some missing features.
[2:05 PM] jflat06: The biggest of the new features is that the Recipes system has been reworked.
[2:05 PM] dudit: @jflat06 When does 64-bit Foldit Windows version come out?
[2:06 PM] jflat06: @Dudit its not currently a priority since it isn't actually actively causing issues. We plan on doing it eventually, but there's not a strong incentive to work on it over our other tasks.
[2:06 PM] alcor29: Can you talk some about the new Recipes system?
[2:07 PM] jflat06: For any of you in discord, I am in the livestream channel (muted) sharing my screen with the new recipe system demo
[2:08 PM] jflat06: The gist of it is that the recipes are now primarily edited through the website.
[2:08 PM] jflat06: And you can maintain a library on the website that will refresh your client recipe list with the click of a button.
[2:08 PM] jflat06: I'll go ahead and make a new recipe now
[2:09 PM] dudit: @jflat06 I think it will be great if we can Pause a recipe
[2:09 PM] HuubR: You mean, pause and do some hand folding, and then continue the recipe where it was paused?
[2:10 PM] jflat06: I'm not quite sure what is involved in that, but it's something we can consider.
[2:10 PM] jflat06: It's sort of out of the scope of the stuff I'm working on right now, though.
[2:10 PM] jflat06: Also included in this, for people who prefer to work locally, you can now load/run a recipe from a raw text file.
NEW
[2:11 PM] LociOiling: you might want to mention that GUI and Lua V1 recipes are going away…
[2:12 PM] pc: @jflat06 In projects the devs know the project well, but not always from a player's point of view. Is the foldit team use foldit often as a player ? We can give you more feedbacks to whatch us play and to better understand our needs, or mistake we do.
[2:12 PM] jflat06: Right, we're retiring GUI/lua v1 recipes as part of this, as it makes the data transfer to the new site much cleaner.
[2:12 PM] Susume: when our recipes live on the server, the scientists can look to see if we have invented any cool algorithms - if we run locally from text file, will foldit team still be able to see the recipe?
[2:12 PM] LociOiling: what happens to GUI and Lua v1 recipes in someone's cookbook? do they disappear, or just not run?
[2:13 PM] jflat06: @Susume they will not, which is why we prefer them on the website. But we thought it was important to include support for local work as well.
[2:14 PM] Susume: ok
[2:14 PM] jflat06: @LociOiling The actual (old) cookbook file will remain on your computer, but they will not appear in the new recipe library.
[2:15 PM] jflat06: We are considering trying to auto-convert old recipes, but it isn't necessarily easy, especially for the GUI ones.
[2:15 PM] Skippysk8sIRC: so will we have to reload our cookbooks again
[2:16 PM] jflat06: @pc Some of us play (more than others). We do make an effort to understand the point of view of the players, though.
[2:16 PM] alcor29: If GUI recipes wont work anymore can you autodelete them from our cookbooks?
[2:17 PM] Skippysk8sIRC: or better still autoconvert lol
[2:17 PM] jflat06: @Skippy You'll configure your new cookbook once on the new site, and then it'll just be a matter of pressing "refresh library" in whatever client you want your library to show up.
[2:17 PM] jflat06: Unfortunately since the idea of a cookbook only exists locally currently, we don't have any way of moving that over to the server.
[2:18 PM] LociOiling: ok, sounds like all.macro is a thing of the past once the new system is up
[2:18 PM] pc: ok. I can do more feedbacks if needed in video). I have some things to show, and there is multiple possible solutions.
[2:18 PM] jflat06: It won't autodelete them, but they will not show up in the new cookbook. You can delete your old all.macro if you want to nuke them off your disc :0
[2:18 PM] jflat06: @LociOiling Yes!
[2:19 PM] jflat06: Recipes are now stored in a 'recipes' folder, and there's 1 file per macro now.
[2:19 PM] jflat06: When you click 'refresh' in the client, it will download any new/changed recipes to make sure you have the latest stuff from the website
[2:20 PM] Skippysk8sIRC: so if we have personal settings we save (e.g. sheet bands sets all bands for sheet bonding) we have to share to public now?
[2:20 PM] jflat06: You'd create a copy on the website. Whether it is shared or not is up to you
[2:21 PM] jflat06: You can make it totally unshared (wont even show up in your own library in-client), self shared, group shared, or public shared.
[2:21 PM] Skippysk8sIRC: kk
[2:21 PM] LociOiling: sounds good so far, avoids copying all.macro all over the place
[2:21 PM] jflat06: We have some neat ideas with creating leaderboards or competitions for recipe writers, but those are all in early development
[2:21 PM] jflat06: But I do want to stress that people who prefer to code locally can still do so - and better than before.
[2:22 PM] jflat06: We decided that an in-client text editor was more of a liability than a feature, so we're just letting you write up the recipes in the editor of your choice, and then simply run the recipe from the file.
[2:23 PM] cjddig: Something went wrong, and I couldn't see the streaming, but question: Will it provide syntax highlighting?
[2:23 PM] jflat06: The website interface does have syntax highlighting
[2:23 PM] jflat06: if you're coding locally, you can use any editor you want.
[2:24 PM] jflat06: (sorry, Discord probably isn't the best place to demo features, especially while I'm trying to chat here at the same time)
[2:24 PM] jflat06: @LociOiling yeah that was a big motivator. It also lets us more accurately track who's using what, so we can get a better idea of which things are good.
[2:25 PM] jflat06: Which also should hopefully help us do a better job at recommending or highlighting recipes to new players
[2:25 PM] alcor29: Will the recipe web listing be more sortable than our cookbooks are now? e.g. I may not want alphbetical order etc. ?
[2:26 PM] jflat06: We're still working on that interface, but I'm thinking it should be.
[2:26 PM] jflat06: There's already a search feature, which is much nicer.
[2:27 PM] alcor29: great
[2:28 PM] dudit: @jflat06 Will the recipe sorting include the most downloaded / most used recipe?
[2:29 PM] donuts554: Will images be able to be used in the Foldit Website Forum and Feedback posts?
[2:29 PM] jflat06: We want to provide multiple ways of ensuring players are finding the recipes that are most useful to them. Most downloaded/used are pretty good metrics, so those will probably be included
[2:29 PM] alcor29: Most correlated with high ranking would be super.
[2:30 PM] jflat06: @Donuts currently on the test site I don't think there's a direct ability to do that, but I am looking into being able to include them more easily.
[2:30 PM] pc: @jflat06 Do we have somes news about Mers puzzles lab tests ? Bkoep talked about soon results there is some months. If it didn't worked, do we know why ? (it is useful for us to adapt how we play)(edited)
[2:31 PM] jflat06: I'll leave that to bkoep as he's more informed about the current state of those results. I will say that the actual physical lab part of biochemistry is a very long process.
[2:32 PM] pc: thanks
[2:32 PM] jflat06: We try and make players aware of results when we think we have a reasonably certain answer to give
[2:32 PM] pc: we are interested by why it didnt work too ^^
[2:33 PM] jflat06: yup! definitely
[2:34 PM] dudit: @jflat06 Will the recipe sorting include a category (example: Early; Early-Mid; Mid; Mid-End; End) ?(edited)
[2:34 PM] jflat06: @Dudit Not currently, but I'm hoping to add a 'tag' system that could be used for that, and more
[2:35 PM] jflat06: Some of these features may come after the fact. Right now we're focusing on the core systems that we cant easily change once things are deployed.
[2:35 PM] alcor29: Most correlated with low rankings also useful.
[2:36 PM] jflat06: @alcor29 a cool idea would be making the recipes and associated data available to everyone, so people could do their own sorting
[2:36 PM] jflat06: But I just came up with that idea right now, so no promises
[2:36 PM] alcor29: Yup
[2:37 PM] Skippysk8sIRC: Folding styles vary — I'm not sure that all recipes fit a game stage nor that we want to direct players to use them that way. Sometimes different approaches can yield equally good results from different processes
[2:37 PM] jflat06: One of my areas of interest is in taking player-generated content and finding the most effective way to filter it to show only the relevant stuff to other players.
[2:37 PM] jflat06: Lots of games have this problem, and lots of them have some not-so-great systems for dealing with it
[2:38 PM] donuts554: But I think that some Foldit Players could maybe skew the recipes to have them be unfairly correlated with low rankings
[2:38 PM] jflat06: @Skippy I agree. It would probably end up functioning something like Steam's tag system, where it is a product of community consensus
[2:38 PM] LociOiling: will links to the old website still work, getting redirected to the new?
[2:38 PM] jflat06: @LociOiling Yeah, that's something that we've talked about, and I think it's handled.
[2:39 PM] jflat06: There might be some links that dont have corresponding pages in the new website, though.
[2:39 PM] alcor29: Results (rankings) seem to be the most important and objective measure.
[2:39 PM] cjddig: How about suggestion feature, that analyzes the recipes someone used and suggests the recipes similar to them?
[2:40 PM] pc: @jflat06 Perhaps the frequently used "recipes" could become tools in GUI. I had thought for example of functions like "mutate" which would be configurable with a UI to add options (like a white list, or special rules)
[2:40 PM] jflat06: @cjddig I think there's actually people working on similar systems in some of the other games we make. It's definitely an idea we've talked about.
[2:41 PM] jflat06: @pc Yup, that's an idea we've talked about, too.
[2:41 PM] jflat06: I'd love for players to be able to add first-party buttons for third-party recipes.
[2:41 PM] pc: player often use recipes instead of hand foldit and expect recipe will do most of the job. Maybe some faster task recipes can encourage more interactivity too ^^.
[2:41 PM] jflat06: (I'll be right back in one minute)
[2:42 PM] donuts554: Why are five Foldit Contests shown to have a Start Date of January 1st, 1930?
[2:43 PM] pc: one of the drawbacks of recipes is that they seek a final solution. Maybe we also need to calculate intermediate solutions, like proteins with only core mutation at the beginning for example.
[2:43 PM] jflat06: I think that date is our "far in the future" date
[2:43 PM] jflat06: Not sure about those specifically
[2:43 PM] pc: (I though of it for linker puzzles for exemple)
[2:44 PM] dudit: @jflat06 I think Batch Running Recipes (Turning Recipes into Actions) will be a great feature
[2:45 PM] Skippysk8sIRC: @Dudit I think Constructor can do that now
[2:45 PM] Skippysk8sIRC: and it is a good idea once hand folded
[2:46 PM] cjddig: Is there any special reason that the programming language of recipe is Lua?
[2:47 PM] jflat06: I think lua is one of the more easily integrated languages for games seeking a scripting language accessible to players.
[2:47 PM] jflat06: But the decision to use lua actually predates me, so I can't say for sure
[2:48 PM] pc: Yea, I have multiple ideas for that ^^(edited)
[2:48 PM] jflat06: I think there's some pretty exciting stuff we can do for showing/recommending recipes to players, especially with the new data about who's using what
[2:49 PM] pc: it can be a simple "graphic programming", but there is more adapted solutions I think.
[2:49 PM] jflat06: But aside from recipes, there's also some other cool features coming with the new site.
[2:50 PM] alcor29: Movable windows?
[2:50 PM] jflat06: One is that when a scientist like bkoep wants to talk about one of your solutions, instead of taking a screenshot and uploading it and trying to describe some feature of a player model… we can now show you the 3D model of any solution in the browser with a link!
[2:50 PM] maithra: @jflat06 Can you give an example how to recommend recipe for symmetric design/electron density/ revisiting type of puzzles?

[2:51 PM] pc: the most important is to identify when players have difficulties, and create tools for that. In design puzzles : create the first structure and make sure it fold with a good core is not so easy. So all player make a tripple helix. off course it is difficult to have large contact zone without buns and not too much hydrophobics.
[2:51 PM] Susume: ooo, can we spin the 3D model you are showing us?
[2:51 PM] jflat06: @maithra I think that's going to come down to a system for evaluating the recipes usefulness in certain contexts
[2:51 PM] jflat06: @Susume yup!
[2:51 PM] HuubR: Cool!
[2:51 PM] cjddig: I'd better turn on my computer… I wanted to see it.
[2:52 PM] pc: oh nice
[2:52 PM] dudit: Great!
[2:52 PM] jflat06: It's a really neat system, and we might expand it in the future to possibly even let players share links to their designs on the website.
[2:52 PM] Susume: @cjddig not showing us right now, it is coming in the new website
[2:52 PM] jflat06: The biggest issue with the system is that it makes it super easy to grab the raw PDB if one is so inclined.
[2:53 PM] jflat06: So to start with, we are going to carefully limit when/where it is used.
[2:54 PM] Susume: yeah, exposing pdbs is suboptimal
[2:54 PM] alcor29: Will the top designs be showed in 3D?
[2:54 PM] jflat06: Unfortunately with the javascript 3D viewers, they need the raw PDB on-hand in the browser in order to display it.
[2:55 PM] jflat06: @alcor29 we are still figuring out the details, but that may be a possibility we consider.
[2:56 PM] jflat06: We've typically been very strict about limiting the amount of info people can put into or pull out of the game, so this is a bit of a change in tune for us, and we're trying to make sure it's all still consistent with our mission and goals.
[2:56 PM] donuts554: What is one thing that you encourage Foldit Players to do, and what is one thing that you discourage Foldit Players to do?
[2:56 PM] alcor29: It would have considerable heuristic value.
[2:56 PM] jflat06: The biggest one is… make more designs!
[2:57 PM] HuubR: Would it be possible (and, would it help) to make a sort of copy-protect feature in Foldit that will recognise a known PDB structure, and exclude it from the competition?
[2:57 PM] alcor29: Pictures and videos can teach a lot more than words.
[2:57 PM] jflat06: Our recent competition that bkoep ran had some super interesting results, and we're looking to implement it as a more formal system with the new site.
[2:58 PM] jflat06: And on that same note, something we discourage (at least scientifically) is heavy drilling, as it isn't terribly useful to use.
[2:58 PM] jflat06: @HuubR for sure, but that is a very hard problem. Especially for prediction puzzles where the whole point is to find the one native solution
[2:58 PM] jflat06: @alcor29 I agree!
[2:59 PM] Skippysk8sIRC: but "heavy drilling" for team competition is some of how we players have fun in last day. Diversity is more likely to start with hand folding at start – fun from taking a team's best design and going for it last day or two
[2:59 PM] jflat06: That was part of my inspiration for trying to move more of the recipe stuff to the website as well. I want players to be able to share their recipes with a direct link to a website page where they can show each other their code and have conversations about it.
[2:59 PM] Skippysk8sIRC: points can be motivating, if not of scientific value
[3:00 PM] jflat06: @Skippy yup, I agree. That's definitely something we take into consideration.
[3:00 PM] pc: I suppose new metrics will be added to design puzzles ?(edited)
[3:01 PM] pc: we are still uncertain whether the protein folds well in lab.
[3:02 PM] jflat06: Probably. We are getting to the point where a lot of the top designs coming out of foldit are looking pretty good, so it is starting to become more of an issue of coming up with lots of them instead of improving them.
[3:02 PM] jflat06: so are we! its a very hard problem. Even the scientists making them using their own software have a pretty low hit rate for designs that fold well in the lab.
[3:02 PM] jflat06: Which is part of the reason that having a large variety is very important.
[3:02 PM] alcor29: Will the newfoldit be faster?
[3:03 PM] jflat06: the new site? I think so
[3:03 PM] donuts554: That is good
[3:03 PM] jflat06: I've done some stress testing, but you never know until you actually deploy it
[3:03 PM] dudit: @jflat06 I think every Foldit puzzle should be a competition puzzle
[3:04 PM] pc: alpha fold metric soon (i hope )(edited)
[3:04 PM] cjddig: Newsletter page takes so much time to load all images. I hope this would be improved.
[3:05 PM] donuts554: Me too
[3:05 PM] jflat06: not quite alphafold, but we have our own version of a machine learning algorithm that we're looking to include as a metric that we hope will be very useful!
[3:05 PM] dudit: @jflat06 With a realtime total number of puzzle submitted
[3:05 PM] pc: oh interesting !
[3:06 PM] jflat06: @Dudit We're not quite sure about for every puzzle, but we are planning to have a full-featured system for these competition puzzles.
[3:06 PM] jflat06: With realtime results!
[3:07 PM] alcor29: Ah using AI algorhythm to solve buns and loops would be great!
[3:08 PM] jflat06: I think it would probably be a supplement to those (especially since it is likely to be quite a bit slower than them)
[3:08 PM] jflat06: but we're still figuring out to best integrate it, and there are some technical hurdles as well.
[3:10 PM] donuts554: Oh ok
[3:11 PM] alcor29: Thanks Jflat. Informative meeting.
[3:12 PM] dudit: Thank you very much @jflat06
[3:12 PM] jflat06: But in any case, you'll all have to thank @Susume for doing a lot of the leg work to take the underlying Rosetta machine learning algorithm and make it intelligible to Foldit players!
[3:12 PM] HuubR: Three cheers for @Susume!
[3:12 PM] dudit: Thanks so much @Susume
[3:12 PM] alcor29: Yes. Thanks @Susume.
[3:12 PM] donuts554: Thanks Susume!
[3:12 PM] pc: thanks
[3:13 PM] Susume: glad to do it!!
[3:13 PM] jflat06: Thanks for the great questions, everyone I am going to take off and get a proper lunch, now.
[3:13 PM] cjddig: Thanks both jflat06 and Susume, time to do my homework…
[3:13 PM] Susume: thanks jflat
[3:13 PM] pc: thanks jflat

[agcohn821]

[2:30 PM] beta_helix: Happy birthday! We'll start the birthday Science chat in a couple minutes
[2:30 PM] Ridick051: oh 13th thanks god it's no FRIDAY
[2:30 PM] Dudit: Hi @beta_helix
[2:31 PM] Ridick051: ok hello @beta_helix
[2:31 PM] Ridick051: ^^
[2:32 PM] milkshakeiii: Woo
[2:32 PM] jflat06: Hello everyone!
[2:32 PM] bkoep: Hi all!
[2:32 PM] Ridick051: hi ^^
[2:32 PM] Dudit: hi @milkshakeiii @jflat06 @bkoep
[2:32 PM] maithra: is it question time yet?
[2:33 PM] beta_helix: YES! Hi everyone! We want to start by thanking you all for coming today, and thanking those who couldn't make it to the party. Foldit would never had made it 13 years without you
[2:33 PM] beta_helix: Fire away maithra
[2:33 PM] maithra: I copied over the previous AA and structure information from the previous 1971 symmetric Tetramers - Hbond networks into the current 1989. I used wiggle and shake, but I never got even near to the shape that I had in the earlier version. What should I do to get the similar shape of the protein when aa and structure are the same?
[2:34 PM] maithra: residue number is the same
[2:34 PM] milkshakeiii: in some puzzles, you are allowed to load solutions from previous puzzles
[2:34 PM] milkshakeiii: but in other puzzles, we want people to start from scratch
[2:35 PM] sethcooper: hello!
[2:35 PM] Dudit: hi @sethcooper
[2:36 PM] ZeroLeak7: hi @sethcooper
[2:36 PM] milkshakeiii: hi @sethcooper
[2:37 PM] rmoretti: Maithra, one thing to keep in mind is that the energy landscape is rugged. Even if you're "close", wiggle and shake may not take you to where you want to be. You may need to use a number of rubber bands to encourage the structure to go where you want it to.
[2:37 PM] josh: hi @sethcooper
[2:38 PM] jflat06: Also note that we don't always want you to re-create an existing fold - we prefer new folds - especially on Design puzzles. Diversity is important, because we need not just good folds, but many of them.
[2:38 PM]
maithra: thank you, rmoretti. What will be the rubber band in the wet lab when you are trying to create the protein? These proteins should fold up without extra help, right?
[2:38 PM] ZeroLeak7: sooo many work with rubber bands and freezing to hold the structure and wiggle and shake etc.
[2:39 PM] beta_helix: Everyone, feel free to fire off a question as there are many of us that can answer in parallel
[2:40 PM] rmoretti: The issue is that Foldit is not a perfect representation of what happens in the wet lab. Rubber bands don't correspond with anything in the wet lab, but rather they're our attempt to overcome the limitations in Foldit.
[2:40 PM] horowsah: @maithra in theory, yes it is going to do what it's going to do in wet lab, but there are things that can be done to influence whether it folds properly or not
[2:41 PM] Susume: @beta_helix I was hoping you could say something about recent ED puzzlses but I guess there is still one open, maybe it's too soon
[2:41 PM] HuubR:In my mind, rubber bands is what you use to get the backbone to the shape you want. Then you use Mutate to fit the AAs that go well with that backbone shape. Does that make any sense at all?
[2:41 PM] Dudit: What is the Priority in Foldit Development & what is the Final Goal in Foldit Development?
[2:41 PM] rmoretti: HuubR - that's a great way of looking at it.
[2:41 PM] ZeroLeak7: is it possible in future, that we can print the proteins with a 3D printer? Bio-printer!
[2:41 PM] HuubR::-) @ZeroLeak7
is it possible in future, that we can print the proteins with a 3D printer? Bio-printer!
[2:41 PM] sethcooper: we've done that in the past!
[2:41 PM] beta_helix: @Dudit The final goal is a Nobel Prize for Foldit players
[2:42 PM] rmoretti: Dudit - There really isn't a final goal in Foldit development – we want to keep improving it as a platform for biochemistry and structural biology research.
[2:42 PM] maithra: My point is: if you want the protein to fold up in the wet lab then it should be able to fold up without rubber band. That was the test I was trying in reloading the same aa and structure from the previous puzzle.
[2:42 PM] ZeroLeak7: woow
[2:42 PM] ZeroLeak7: Nobel Prize =.O
@Dudit
What is the Priority in Foldit Development & what is the Final Goal in Foldit Development?
[2:42 PM] neilpg628:I don't know if there is a final goal (others might have a different opinion) but I think the priority is to find and solve problems in protein design that human ingenuity is specifically useful for
[2:42 PM] rmoretti: As long as there's unsolved problems citizen scientists can contribute to, we want to make that possible.
@Dudit
What is the Priority in Foldit Development & what is the Final Goal in Foldit Development?
[2:43 PM] sethcooper: one thing we are looking at is making the tutorial more adaptive to how well players do
[2:43 PM] maithra: I would like to know whether the foldit designs you have successfully recreated in the past were able to fold without any help.
[2:43 PM] sethcooper: i think we're also interested in a VR version
[2:43 PM] Susume: rubber bands help it get into the shape you want, then you have to remove rubber bands and get it to stay there and score well
[2:43 PM] milkshakeiii: in terms of science we have several teams working on different goals, but some of the important ones are: binders, symmetry puzzles, electron density
[2:44 PM] jflat06: @maithra - one thing to note is that foldit isn't a physics simulation. Wiggle isn't a physics simulation - it is just an optimization algorithm that is relying on physical and statistical score terms.
[2:44 PM] rmoretti: Maithra - True. A good step is to remove the rubber bands and see if the structure is stable without them. – And all of the foldit designs which work in the lab have been able to fold without help (or at least without any more help than is normally present in cells.).
@Susume
rubber bands help it get into the shape you want, then you have to remove rubber bands and get it to stay there and score well
[2:44 PM] ZeroLeak7: yes, but it is not that easy
[2:44 PM] jflat06: So we wouldn't expect it to be able to find shapes alone - it needs things like rubber bands to guide it towards the true shape, with the assistance of a human brain
[2:45 PM] ZeroLeak7: I have try to bend the helices now @Susume
[2:45 PM] Susume: you can get the 3d printer plans for the foldit designs in the PDB from: https://3dprint.nih.gov/
[2:46 PM] brunokestemont: HappyFoldit brirthay all. OK for Player's Nobel price ;)
[2:47 PM] Sciren: Yes indeed happy birthday Foldit!
[2:48 PM] horowsah: I wonder if you can design a protein birthday cake…
[2:48 PM] brunokestemont: In a far future, they will think Foldit is the First name of Players. Foldit will become a commun first name.
brunokestemont: In a far future, they will think Foldit is the First name of Players. Foldit will become a commun first name.
[2:49 PM] beta_helix: https://pubmed.ncbi.nlm.nih.gov/?term=foldit+players
PubMed
foldit players - Search Results - PubMed
foldit players - Search Results - PubMed

[2:49 PM] milkshakeiii: alpha helixes would be pretty hard to make into a cake xD
[2:50 PM] maithra: Another question: in which part of the cell is the protein folding up in the wet lab? In the cytoplasm or in the nucleus of the cell?
[2:50 PM] Susume: oh sad, pubmed does not find our ED papers under our name
[2:50 PM] rmoretti: People were showing some pretty impressive barrel structures that would probably make a nice cake.
[2:51 PM] alcor29:Alternate layers of bacon and eggs.
[2:52 PM] HuubR:Donut? (hang on for a screenshot…)
maithra: Another question: in which part of the cell is the protein folding up in the wet lab? In the cytoplasm or in the nucleus of the cell?
[2:52 PM] neilpg628: That would be in the cytoplasm
[2:52 PM] rmoretti: @maithra Proteins aren't typically expressed in the nucleus. Proteins are almost always expressed in the cytoplasm, and then moved to the nucleus later. (Also, they're often expressed in E. coli bacteria, which don't have a nucleus at all.)
[2:52 PM] HuubR:IMAGE: http://fold.it/portal/files/chatimg/irc_977284_1620679955.png
[2:52 PM] ZeroLeak7: @HuubR wooow
[2:52 PM] rmoretti: (Moved to the nucleus later if they need to be in the nucleus – most proteins don't.)
[2:53 PM] Susume: #beta_bagel
[2:53 PM] HuubR:lol, I like that term :-)
[2:53 PM] jflat06: yeah that's great
HuubR:IMAGE: http://fold.it/portal/files/chatimg/irc_977284_1620679955.png
[2:53 PM] Sciren: #beta_bundt_cake

[2:53 PM] horowsah: there we go!
[2:53 PM] donuts554:hello
[2:54 PM] beta_helix: speaking of donuts…
[2:54 PM] ZeroLeak7: donuts554 hello
[2:55 PM] donuts554:Hello how are you?
[2:55 PM] brunokestemont: Donuts by HuubR scores well. i wonder: it is possible in nature ?
[2:55 PM] beta_helix: @everyone Any questions about the 13 years of Foldit?
[2:56 PM] jflat06: we have quite a few foldit historians here
[2:56 PM] brunokestemont: Do you have an ie of the total number of players who at least tried to play som epuzzles ?
[2:56 PM] horowsah: on the question of whether you can make a protein bagel in nature
[2:56 PM] brunokestemont: idea
[2:56 PM] ZeroLeak7: is there a list, how many people worked with Foldit: Developers, Scientist etc.?
[2:57 PM] maithra: Do you have plans to move to/ create a 64 bit version of Foldit? Most PC-s come with 8-16 GB memory nowadays - what cannot be used all by a 32-bit foldit.
[2:57 PM] ZeroLeak7: in Numbers
@ZeroLeak7
is there a list, how many people worked with Foldit: Developers, Scientist etc.?
[2:57 PM] josh: https://fold.it/portal/info/credits this?
[2:57 PM] sethcooper: I think it's well over 500K but I haven't checked in a while
@beta_helix
@everyone Any questions about the 13 years of Foldit?
[2:57 PM] Dudit: Are there any prediction of when will Foldit out of Beta Version?
[2:57 PM] ZeroLeak7: woooow
[2:57 PM] HuubR:Wow, @Susume :-o
[2:57 PM] ZeroLeak7: thank you seth
@Dudit
Are there any prediction of when will Foldit out of Beta Version?
[2:57 PM] josh: Technically, never. We are a "continuous beta" software
[2:57 PM] ZeroLeak7: good to know
@Dudit
Are there any prediction of when will Foldit out of Beta Version?
[2:58 PM] josh: https://en.wikipedia.org/wiki/Perpetual_beta
Perpetual beta
Perpetual beta is the keeping of software or a system at the beta development stage for an extended or indefinite period of time. It is often used by developers when they continue to release new features that might not be fully tested. Perpetual beta software is not recommended for mission critical machines. However, many operational systems fin…
[2:58 PM] donuts554:What was done on Foldit's first birthday?
[2:59 PM] HuubR:@josh: in other words: it's not a bug, it's a feature
[2:59 PM] MirsadaH:Hi everyone, is there any chance to see Foldit app on Android? If, yes, where is it in store?
[2:59 PM] josh: ;)
MirsadaH:Hi everyone, is there any chance to see Foldit app on Android? If, yes, where is it in store?
[2:59 PM] sethcooper: there is a chance
[2:59 PM] ZeroLeak7: yes android, would be nice!
[3:00 PM] MirsadaH:nice, looking forward to see it soon
[3:00 PM] brunokestemont: I see a new user with number 1030585. Would it be near to the total number of players (counting the mysterious Ashley)?
[3:00 PM] ZeroLeak7: would be awesome mobile folding hehe
[3:01 PM] beta_helix: Doesn't Ashley have more than one account?
[3:01 PM] jflat06: @MirasadaH there's some technical hurdles to releasing on android/iOS, but we're working on some things that could make it easier
[3:01 PM] Susume: Ashley is legion
[3:01 PM] josh: brunokestemont that might be a rough approximation
[3:01 PM] ZeroLeak7: it should run on nvidia shield
[3:02 PM] ZeroLeak7: streaming is the future
[3:02 PM] alcor29: How about fire OS?
[3:03 PM] Dudit: Make Foldit touchscreen friendly
@Dudit
Make Foldit touchscreen friendly
[3:03 PM] sethcooper: Foldit should be touchscreen friendly!
[3:03 PM] MirsadaH:or even compatible with VR hardware/software hehe
[3:04 PM] Formula350: I don't think that number is accurate Bruno. Baker's user ID is something like 2430, and from what I could tell that was the first user.
[3:04 PM] sethcooper: we at one point had a touchscreen version but I'm not sure if it made it's way into the released version

[3:04 PM] donuts554:What do you think is Foldit's most memorable birthday?
[3:04 PM] sethcooper: a few things had multitouch support - pull, move, and camera controls mainly
[3:05 PM] horowsah: well, on the 21st birthday, maybe we'll do a puzzle on alcohol dehydrogenase
[3:06 PM] Formula350: The Android Remote Control app has limited touch capability, as does Foldit in general. It's not the most friendly for input/control, but it IS doable. I think of the Remote Control app as more of a thing to check up on how a recipe is running or start a new one. I don't know if it'd handle actual play very well. (But I hadn't tried, eitehr!)
[3:06 PM] beta_helix: @donuts554 I just looked to see if we did anything for the first Foldit anniversary, but I don't see any posts about it. I only found an email from sethcooper sent to the Foldit team (which was probably only 3 people at the time!)
[3:07 PM] Formula350: Though I'd love to see the Android App (since it's just a Unity-Engine front-end) ported to Desktop, to let us run it as the front-end UI for various clients opened either on the same computer or via network. Allowing us to jump betwixt the clients from a single remote window :}
[3:07 PM] sethcooper: Foldit user id 1 was originally Janos I think
[3:08 PM] jflat06: (since re-written to the Admin account)
[3:09 PM] maithra: A desktop/laptop computer with 4 cores might run 4 different clients with Formula's design idea. I would love that.
donuts554:What do you think is Foldit's most memorable birthday?
[3:09 PM] beta_helix: Probably the 5th anniversary, because katfish made a really cool poster for it… with a timeline of all your accomplishments.
[3:09 PM] Formula350: Janus was the roman deity of doorways. A "Janus Account" would make for a cute nickname for a Admin Account, in that regard haha
[3:10 PM] sethcooper: i think the next that's an actual person is Zoran with 42
@Formula350
Janus was the roman deity of doorways. A "Janus Account" would make for a cute nickname for a Admin Account, in that regard haha
[3:10 PM] Sciren: That's actually super clever!
[3:10 PM] donuts554:Oh ok
[3:11 PM] jflat06: Meanwhile I'm sitting here at 220130
[3:11 PM] SuperNova185: This is a back-in-the-day blog link to FoldIt MacArthur grant appeal: https://www.nanopaprika.eu/profiles/blogs/support-stem-gaming-support
The International NanoScience Community - Nanopaprika.eu
Support STEM Gaming / Support FOLDIT's MacArthur Foundation grant a…
. The Foldit research team (read my earlier post) is striving to extend the Foldit application into a biochemistry tool for grade-schoolers [http://fold.it/portal/] in order to "develop specific technological and social scaffolds to support learning ofmultiple science proficiencies (e.g., interest in science, abstractknowledge, scientific prac…
[3:12 PM] Dudit: Is it possible that Foldit computation runs on the server (to minimize the lag) ?
[3:12 PM] SuperNova185: This is when I was an academic librarian trying to get STEM in the programmes. The powers that be went with STEAM instead.
[3:12 PM] ZeroLeak7: coool
[3:12 PM] Formula350: Maithra that's pretty much what I was envisioning, yes My hope being that with the client minimized and the Remote Control app, that it might not experience the same "slowdowns" while running a recipe since that client could perhaps have its UI stuff offloaded to this Remote Client. However that's also my non-coder-dray-dreaming at play… lol
beta_helix
Probably the 5th anniversary, because katfish made a really cool poster for it… with a timeline of all your accomplishments.
[3:12 PM] beta_helix: sadly I lost the physical copy of the Foldit poster coming back from the Science & Engineering festival in DC… Southwest never found it. I like to think there is someone near the Providence airport with a cool Foldit poster in their room
[3:12 PM] Formula350: hah Thanks Sciren
[3:12 PM] SuperNova185: I interviewed Seth Cooper for my nanogaming blog.
[3:13 PM] jflat06: @Dudit if you pay for it! It's actually not very practical to do that, because we would end up running the computations for an unbounded number of clients.
[3:13 PM] SuperNova185: That was 2010.
[3:13 PM] jflat06: Which would grind our servers to a halt pretty quickly after even a handful of people were connected.
[3:13 PM] ZeroLeak7: I would pay 9,99 $
[3:14 PM] ZeroLeak7: in month
[3:15 PM] SuperNova185: This was when FoldIt was competing for the Digital Media Learning funding grants : hey've entered the Digital Media Learning Competition [http://www.dmlcompetition.net/pligg/story.php?title=849
[3:15 PM] SuperNova185: as contenders for a Learning Design Labs award, underwritten by the MacArthur Foundation.
[3:15 PM] maithra: I thought the servers only pass the scores of other players. The folding happens on the client computer, right? Passing scores cannot be such a demand on the server.
[3:16 PM] Formula350: Yes Maithra. I think Dudit was pitching a request/idea to have a Server-Side version, to perhaps allow for something like a Mobile version, so that your mobile device wouldn't be taxed with the computational workload.
[3:16 PM]
donuts554:Oh ok
[3:16 PM] ZeroLeak7: can we spend money to Foldit?
[3:17 PM] maithra: Our mobile devices have quite a powerful computer in them, I think.
[3:17 PM] SuperNova185: So, can I regain all of my points garnered over more than a decade - lost during hiatus and sabbatical?
[3:17 PM] maithra: Our smartphone can do face recognition - that is AI. I expect it could handle the folding as well.
[3:18 PM] donuts554:How could the protein folding problem be solved?
[3:18 PM] beta_helix: Other games (such as our friends at EteRNA https://eternagame.org/about) do take donations, but you all contribute so much already that we avoided this.
[3:18 PM] ZeroLeak7: exactly, but our computers could handle much more AI too. We need new functions to write AI-RECIPES
[3:19 PM] Formula350: I suspect they'd be capable of folding if it were ported to Android/iOS, in so far as being not much slower than doing it on a PC with a CPU running at a similar frequency (GHz). Though, that comes back to the touch interface and whether that's sufficient enough…
[3:19 PM] jflat06: mobile processors are definitely getting more powerful, but they still have to operate on very small power budgets that keep them from competing with desktop processors
[3:19 PM] maithra: I would rather have a 64-bit version of foldit than a mobile version.
[3:19 PM] brunokestemont: For AI, don't yo need access to huge databanks? I s rebuild or remix a kind of AI?
[3:20 PM] sethcooper: 64 bit does seem pretty reasonable!
[3:20 PM] Formula350: Granted they lack some instruction sets, which Foldit may rely heavily on, so if they're being emulated on ARM that might be problematic.
[3:20 PM] ZeroLeak7: neuronal networks is a big thing @Bruno Kestemont
[3:21 PM] Formula350: We could run AI via OpenCL on our GPU's cough lol But that's a WHOLE other bag of worms lol
[3:21 PM] rmoretti: brunokestemont - It depends on the type of AI, but a lot of the newer ones require large datasets in the training phase. Typically you don't need to have access to the databanks for the actual usage, as the info in the datasets are embedded (in a sense) in the model themselves.
[3:21 PM] ZeroLeak7: yeees thats it @Formula350
[3:21 PM] jflat06: Rebuild/Remix aren't AI - rebuild just takes known segments of 3-residues and inserts them to make a new backbone for a region… and then tries to make it reconnect. Remix does spatial hashing to do a lookup for known good backbones that have the correct endpoints to hook up with your two chosen endpoints.
[3:22 PM] ZeroLeak7: we could train the neuronal net with Foldit for a few years
[3:23 PM] ZeroLeak7: some Ideas ^^
[3:23 PM] rmoretti: We are interested in incorporating AI techniques into Foldit, though. trRosetta (the Baker Lab's verison of Alpha Fold) is the obvious first step, but other places we could employ AI to help out players is something that's on our minds. (Many in the Rosetta community are exploring various ways of applying AI to structural modeling tasks.)
[3:23 PM] maithra: foldit users can generate many false solutions so far. AI needs positive examples too. I mean examples that are folding up in the wet lab properly.
[3:25 PM] maithra: I think that the 32 bit version of foldit is not up to AI tasks. You need 64 bit foldit for that.
[3:25 PM] Formula350: We need a new tool: "Re-re-re-REMIIXX airhorn sound" which lets Rebuild run until it finds a suitible pose, and then Remix finish it off by inserting valid ("known good backbone") poses. If the Remix portion is unable to, then it tries again with a new Rebuild->Remix. Might allow for Remix to work on sections >9seg like Rebuild can, if it broke up the size into digestible chunks.
[3:25 PM] ZeroLeak7: yes that's not easy, but we can sort out the false solutions
[3:25 PM] ZeroLeak7: yes we need 64 bit a lot of ram for AI
[3:25 PM] pc: hi
[3:25 PM] pc: happy birthday ^^
[3:26 PM] beta_helix: @pc Hello, welcome to the party!
[3:26 PM] ZeroLeak7: A 64-bit register can theoretically reference 18,446,744,073,709,551,616 bytes, or 17,179,869,184 gigabytes (16 exabytes) of memory. This is several million times more than an average workstation would need to access
[3:27 PM] ZeroLeak7: soon we have exascale supercomputers in 20 years hehe
[3:27 PM] jeff101: Happy Birthday to Foldit! I'm glad to see experiments are running in the lab again.
[3:28 PM] beta_helix: Any final questions?
[3:28 PM] Formula350: I might be wrong but I think the benefit in Foldit's context would be that 64bit registers could allow faster transfer of data vs 32bit… but I'm venturing outside my area of even partially knowing what I'm talking about lol
[3:28 PM] beta_helix: @jeff101 so are we
[3:28 PM] ZeroLeak7: hmm ok so it will be
[3:28 PM] Formula350: Is the Cake also a lie in Foldit?
[3:28 PM] HuubR:?
[3:28 PM] alcor29:What do you all see for this coming year?
[3:29 PM] ZeroLeak7: fun
[3:29 PM] maithra: 64 bit computer programs can use more memory than 32 bit computers - that is the big difference.
@Formula350
Is the Cake also a lie in Foldit?
[3:29 PM] Sciren: The cake is always a lie, you should never trust it.(edited)
[3:29 PM] ZeroLeak7: have fun
[3:29 PM] pc: is David Baker connected to the chat ?
[3:29 PM] jflat06: except it wasn't a lie in the end!
[3:29 PM] Formula350: (That's a Portal game reference for anyone not familiar, just google "The cake is a lie")
[3:29 PM] HuubR:Thx
[3:29 PM] ZeroLeak7: Living you Life Guys and Girls
[3:30 PM] beta_helix: @alcor29 I see exciting Electron Density results and improvements
[3:30 PM] ZeroLeak7: cya
[3:30 PM] Formula350: Cya Zero
[3:30 PM] Formula350: Do we see a new website coming online? glances at jflat
[3:30 PM] horowsah: goodbye all, i look forward to protein cake!
[3:30 PM] sethcooper: bye everyone! thanks!

[agcohn821]

4:00 PM] milkshakeiii: Hello! I am a programmer at the IPD where Foldit was originally created, and my office hours is starting now. I can talk about the technical side of Foldit.
[4:00 PM] milkshakeiii: Alternatively, I can talk about why Illinois is the best state in the USA
[4:01 PM] milkshakeiii: (Because corn is delicious, duh)
[4:03 PM] Susume: lol hi milkshake!
[4:04 PM] Susume: I'm only here for a couple minutes, but could you talk a bit about how trRosetta is going to be exposed in foldit?
[4:04 PM] jmbrownlee333: Escape to Wisconsin
[4:04 PM] milkshakeiii: I could talk about that, however I think you are the expert on it!
[4:05 PM] milkshakeiii: I believe we are thinking about trying to make it a simple metric at first
[4:05 PM] Susume: the part I know nothing about is the user interface
[4:05 PM] milkshakeiii: Probably a simple objective would be the first step
[4:05 PM] milkshakeiii: Then maybe another window like the blueprint window with information
[4:06 PM] Susume: ok cool! I gotta run, have fun yall!
[4:06 PM] milkshakeiii: Ok see ya!
[4:12 PM] pc: hi
[4:13 PM] milkshakeiii: o/
[4:16 PM] pc: are new lua functions will be added in foldit ?
[4:17 PM] pc: I waited somes like "detect core in a protein" or add a black / white list in mutate.
[4:17 PM] pc: custom mutate in recipes are very slow
[4:18 PM] pc: but of course it can be good if we can add some white list as a player too.
[4:18 PM] pc: and for each segment (not only a global white list)
[4:19 PM] milkshakeiii: we have recently added the ability to retrieve information from filters in lua scripts
[4:19 PM] milkshakeiii: detect core in a protein sounds like something that could be done using that interface
[4:19 PM] milkshakeiii: using the core existence filter
[4:19 PM] pc: ah cool. idealy it is better if thoses fonctionscan be called without filter in the puzzle.
[4:20 PM] milkshakeiii: yeah that would be good. at least core existence filter is pretty common tho I think
[4:20 PM] milkshakeiii: white listing residues during mutate also sounds like a good idea
[4:21 PM] milkshakeiii: i think you can sort of do that using selections?
[4:21 PM] pc: when I create a base protein in design puzzle, when I mutate I have often serine in core.
[4:21 PM] pc: there is some tips to avoid that, but it is better if the first mutate give good AA at begin ^^
[4:22 PM] pc: I can use core existence with the new lua fonctions so, it is good ^^
[4:24 PM] milkshakeiii: I see, yeah
[4:24 PM] pc: creating a first protein is not easy. maybe we will need later a new tool to construct them (and not only tripple helix)
[4:25 PM] milkshakeiii: maybe hallucination using trRosetta
[4:25 PM] pc: I often create tripple helix because I have a lot of difficulty to keep a correct sheet in my protein
[4:27 PM] milkshakeiii: from what I understand, protein scientists at the IPD do the same thing, so, you're not alone
[4:27 PM] milkshakeiii: also even nature makes a lot of triple helixes i think
[4:27 PM] milkshakeiii: i'm not the expert on protein science tho so take what I say with a grain of salt
[4:28 PM] pc: Yes. Maybe it can be usefull to try some differents shapes, because some design puzzles can have no very good solutions with a tripple helix.
[4:28 PM] milkshakeiii: yeah diverse backbones are more exciting and interesting
[4:29 PM] pc: I saw Mers puzzle results. Maybe we needed bigger contact surface area (sorry it is a scientific question ^^)
[4:29 PM] pc: players need to know "what to do now" after something didnt work.
[4:30 PM] pc: or "is there a new tool to test something else?"
[4:30 PM] milkshakeiii: yeah, I know bkoep is working on that
[4:30 PM] pc: are some of our solutions tested in a physic simultor to know why it didnt work in lab ?
[4:31 PM] milkshakeiii: usually they aren't tested in the lab unless they look good from the simulation standpoint
[4:31 PM] pc: It can be very intresting for player to see that too. I already saw for a protein fold in 3d from rosetta. it was cool ^^
[4:32 PM] pc: Can be interesting to see when it comes off and why
[4:34 PM] pc:
https://www.youtube.com/watch?v=gFcp2Xpd29I : like this but for design puzzles
YouTube
Pande Lab Science
Simulation of millisecond protein folding: NTL9 (from Folding@home)

[4:35 PM] milkshakeiii: oh, interesting
[4:35 PM] milkshakeiii: also, also sprach zarathustra as the background music ha
[4:38 PM] alcor29: Hi milkshake. What about this dream scenario. We are given x number of residues. Then in a sort of blueprint tool we specify I want y number of helices, sheets, and ferrodoxing fold with proper chirality? It would save us a lot of 'menial' uncecessary work.
[4:39 PM] pc: yea like a super tool that create the first structure ^^
[4:39 PM] milkshakeiii: yeah the blueprint tool now is quite clunky
[4:39 PM] milkshakeiii: just today in a meeting somebody mentioned redoing it
[4:40 PM] milkshakeiii: it needs some iteration
[4:40 PM] HuubR: I often use the Blueprint window just as a tool to quickly select a number of residues.
[4:40 PM] pc: blueprint tool still usefull in midgame sometimes
[4:41 PM] milkshakeiii: yeah it's useful for sure, could be even more so
[4:42 PM] pc: but yes, when a loop become not ideal, it is very hard to fix. remix doesnt work most of the time. blueprint make too much moves.(edited)
[4:43 PM] HuubR: What if we could copy and paste the backbone shape from one part of the Blueprint window to another?
[4:45 PM] pc: I imagine this : A super creator tool, in a "protein creation mode" at start that have no editable residues. Only Helix, loops and sheets that we can move or turn. and when it is ok, we can convert it to regular foldit protein.(edited)
[4:46 PM] pc: and of course create 310 helix
[4:47 PM] pc: There is 2 kinds of Helix if I understand
[4:47 PM] pc: https://en.wikipedia.org/wiki/310_helix
310 helix
A 310 helix is a type of secondary structure found in proteins and polypeptides. Of the numerous protein secondary structures present, the 310-helix is the fourth most common type observed; following α-helices, β-sheets and reverse turns. 310-helices constitute nearly 10–15% of all helices in protein secondary structures, and are typically obser…

[4:47 PM] milkshakeiii: that would be quite great
[4:48 PM] alcor29: What area are you working on currently milkshake?
[4:48 PM] pc: 310 helix are maybe better for symetric puzzles
[4:48 PM] milkshakeiii: i'm currently working on improving the integration between Foldit and rosetta
[4:49 PM] milkshakeiii: Foldit is currently using a modified version of rosetta from 6 years ago, with some parts from 13 years ago
[4:49 PM] milkshakeiii: we want to be able to easily plug in new versions of rosetta as they come along
[4:49 PM] milkshakeiii: so I'm working on that
[4:50 PM] alcor29: So it's a conversion project but not necessarily increasing our design capacities?
[4:50 PM] milkshakeiii: yes, but the new versions of rosetta are needed for a lot of design stuff
[4:50 PM] pc: we will have new "score parts" in this new version of rosetta ?
[4:50 PM] milkshakeiii: such as introducing new calculations or trRosetta
[4:50 PM] milkshakeiii: I think there have been some minor changes in score parts in the last 6 years, yes
[4:51 PM] milkshakeiii: however, the core score function hasn't changed much
[4:51 PM] milkshakeiii: objectives/filters have, though
[4:52 PM] pc: And maybe some score parts will replace some objectives. Mutate function often doesn't take count about objectives (like BUNS)
[4:54 PM] milkshakeiii: yeah the interaction between mutate and objectives is another important question
[4:54 PM] pc: thanks ^^
[4:54 PM] cjddig: Are there the thing you're planning to make/release shortly?
[4:55 PM] milkshakeiii: well, dojo mode just got released!
[4:56 PM] milkshakeiii: yesterday!
[4:56 PM] milkshakeiii: and there is another thing coming up which I'm not sure if it has been announced yet
[4:57 PM] milkshakeiii: it hasn't been, secret secret 8)
[4:57 PM] cjddig: I'm enjoying dojo, but I think there're some technical issues about it.
[4:58 PM] milkshakeiii: yeah there are
[4:58 PM] milkshakeiii: like slow load times between levels
[4:58 PM] milkshakeiii: Josh is hard at work on that
[5:00 PM] pc: most of dojo puzzles are solve with wiggle
[5:00 PM] milkshakeiii: i think once you solve a few easy ones liek that it should start giving you hard ones though
[5:00 PM] cjddig: I think not that many people play dojo now, but it is the fact that it's funny mode. Now I cannot solve most levels I meet!
[5:01 PM] milkshakeiii: yeah Josh needs more data so he can create a range of difficulties as opposed to jsut wiggle/unsolveable
[5:01 PM] milkshakeiii:
so play it more and Josh will have more data ^_^
[5:02 PM] alcor29: Thanks milkshakeiii.
[5:02 PM] milkshakeiii: thanks all! the office hours is technically over now but I'm still around for a bit 8)
[5:02 PM] pc: thanks ^^
[5:03 PM] cjddig: Thanks!
[5:03 PM] pc: thanks for answers.
[5:04 PM] milkshakeiii: Thanks for questions 8)
[5:05 PM] cjddig: I had breakfast in the middle of Office Hour… I couldn't be here.
[5:06 PM] milkshakeiii: the most important meal of the day
[5:08 PM] pc: it is 1:00h in france ^^
[5:09 PM] milkshakeiii: yikes ha
[5:10 PM] alcor29: Do you have an idea when the conversion will be done?
[5:11 PM] milkshakeiii: the first pass will be done next month probably but then there are many more steps after that
[5:12 PM] alcor29: Ah. Tx.
[5:12 PM] alcor29: And I'm sure there will always have to be maintenance.
[5:14 PM] milkshakeiii: there's always that, ha, hopefully this will actually help reduce maintenance though
[5:17 PM] milkshakeiii: i'm off, good luck all!

[agcohn821]

2:00 PM] beta_helix: Yes it does! Seth and I will start it soon
[2:02 PM] beta_helix: Hi everyone, and welcome to Foldit Office hours! Seth Cooper and I will be here for the next hour to answer any questions.
[2:03 PM] beta_helix: We're also happy to talk about the new C++ developer job ad that we posted: https://fold.it/portal/node/2011713
Foldit is hiring: C++ developer focused on Electron Density
Foldit is hiring: C++ developer focused on Electron Density
[2:03 PM] sethcooper: hello!
[2:03 PM] spvincent:hello!
[2:03 PM] jeff101:hi
[2:04 PM] beta_helix: Hi everyone
[2:04 PM] HuubR:Hi from Europe as well :-)
[2:04 PM] beta_helix: I hope it's not too late
[2:04 PM] spvincent:I was wondering about the binder design puzzles: they seem a bit of a disappointment.
[2:05 PM] BletchleyPark: good evening
[2:05 PM] spvincent:How different are the computer-designed solutions that showed some binding in the last mers puzzle.
[2:05 PM] beta_helix: As my thesis advisor used to tell me when I was getting frustrated: "It was an easy problem, it would have already been solved."
[2:05 PM] beta_helix: Hi BP!
[2:05 PM] BletchleyPark: made it o this one :)
[2:05 PM] spvincent::-)
[2:06 PM] beta_helix: Yay!
[2:06 PM] beta_helix: I will be honest and say that Seth and I are not the binder design specialists, so we don't know much more than the blogpost about it: https://fold.it/portal/node/2011554
Experiment results for MERS-CoV binders
Experiment results for MERS-CoV binders
[2:06 PM] jeff101:what did you propose in the NSF / cryo-EM grant that was funded?
[2:06 PM] beta_helix: However, I tend to think about it in terms of where Foldit design was 10 years ago.
[2:07 PM] beta_helix: Seth, remember how it seemed like a bit of a disappointment at the time?
[2:08 PM] sethcooper: well, there was certainly some work to do to get to working designs!
[2:08 PM] spvincent:Perhaps we could have have one of the partially successful designs as a starting point 8in a future puzzle.
[2:09 PM] beta_helix: Exactly. It was really rough for a while. Think back to the early symmetry puzzles where the top scoring solutions looked really cool (like a pillow, for example) but would never fold in nature. We had to fix the score function significantly to get to the 2019 results.
[2:09 PM] jeff101:what can you tell us about Foldit developments coming soon? what are you working on?
[2:09 PM] beta_helix: (and when I say "we", I mean bkoep )
[2:10 PM] sethcooper: i have been working on a 64-bit windows build
[2:11 PM] jeff101:that's good. folks have been requesting that.
[2:11 PM] BletchleyPark: That shouls allow for larger designs
[2:11 PM] sethcooper: i'd say its pretty much working at this point(edited)
[2:11 PM] sethcooper: i may have fixed audio for Linux in the process (I think I heard that didn't work for some people)
[2:12 PM] BletchleyPark: I notice that keyboard commands get lost in the client
[2:12 PM] BletchleyPark: has this been tracked down as well ?
[2:12 PM] BletchleyPark: (and congrats for getting the 64 bit build working)
[2:12 PM] sethcooper: that i'm not sure about
[2:12 PM] jeff101:are folks doing experiments in the lab again?
[2:14 PM] beta_helix: @jeff101 w.r.t the funded NSF grant, it's basically tackling 2 new components:
[2:14 PM] beta_helix: • Integrating new crystallography refinement capabilities within Foldit. • Adding new modes that allow much larger proteins to be handled in Foldit for cryo-EM.
[2:15 PM] HuubR:Sounds interesting, but please tell us a bit more :-)
[2:15 PM] beta_helix: The general idea is to make Electron Density puzzles a lot easier for you… I know jeff101 has suggested many cool ED features, for example
[2:16 PM] jmbrownlee333: I wish ED puzzles handled discontinuous fragments better. not scoring -99999
[2:17 PM] beta_helix: So one goal (specifically for large cryo-EM structures) is to implement a new Foldit mode where rather than all residues being scored, only the subset being worked on by the user at that time would be used.
[2:17 PM] jmbrownlee333: also wish we had poly ala capabilities.
[2:17 PM] jeff101:are one-way bands on your list of features to implement?
[2:18 PM] BletchleyPark: Would the 64-bit version also allow for more AI assistance in the client for design choices ?
[2:18 PM] beta_helix:
The remainder of the protein will have its energy cached before being recombined.
[2:18 PM] jmbrownlee333:nice. yhat will help a lot
[2:18 PM] beta_helix: @jmbrownlee333 yeah, hopefully that will address your concern!
[2:19 PM] sethcooper: it might potentially if the AI needs more memory, so that's a possibility
[2:19 PM] HuubR:So how do you define the subset that is being worked on?
[2:20 PM] BletchleyPark: Could there be a way for clients to communicate with eachother, either through a disk file (Share mode) or otherwise ?
[2:20 PM] BletchleyPark: as in: exchange paramters, results etc
[2:21 PM] sethcooper: maybe, what would the use case for that be?
[2:21 PM] BletchleyPark: I find loading and saving solution files very slow lately, something faster than that would be welcome
[2:21 PM] HuubR:BP, you mean parallel computing?
[2:21 PM] beta_helix: @HuubR You would essentially use cutpoints, but you can imagine this being particularly useful with much bigger density clouds… if we don't know where a monomer fits in a huge density that fits multiple domains, for example.(edited)
[2:22 PM] BletchleyPark: I run several clients, looking for a good pose, if one of them find one, it could communicate it to the other ones, yielding a new startpoint
[2:22 PM] BletchleyPark: @HuubR, essentially yes
[2:22 PM] beta_helix: @jeff101 once we hire someone dedicated to working on this, we plan to go through the long list of ED features that we've wanted to implement over the many years (edited)
[2:23 PM] jeff101:how long can the position last?
[2:23 PM] sethcooper: i see
[2:23 PM] Susume (she/her): so like that giant virus donut, where there were lots of copies, we might have half the protein in density copy A and the other half in density copy B
[2:23 PM] beta_helix: The grant is 3 years.
[2:23 PM] beta_helix: Exactly, Susume.
[2:23 PM] sethcooper: i'm not sure if the file system itself would be the slowest part of loading a solution
[2:24 PM] BletchleyPark: I use SSDs and the clients used to be quicker in loading
[2:25 PM] beta_helix: Or you can imagine there is a big cryo-EM complex with thousands of residues, and we just want to tackle one 300-residue domain in there… but we don't know where it fits. With this feature we could give you that 300-residue chain and you would be able to try all the different places where it could fit (without completely killing your CPU!)(edited)
[2:26 PM] sethcooper: i wonder if the protein size / filters might be part of it as well
[2:26 PM] BletchleyPark: Are the filters enabled while loading ? You might switch that off automatically while constructing the model as it loads
[2:27 PM] BletchleyPark: same for saving
[2:28 PM] sethcooper: i think if the filters are on they get run when loading so the protein can be scored
[2:28 PM] Susume (she/her): @beta_helix do you have any preliminry results from recent ED puzzles?
[2:33 PM] BletchleyPark: test
[2:34 PM] beta_helix: @Susume (she/her) I wanted to look up the correct details, because I know they gave us another target that we will be posting soon. For the 2 previous cases they were able to build the structure by comparing to previous vWa domain crystal structures.
[2:35 PM] Susume (she/her): sry what is vWa?
[2:36 PM] beta_helix: We are fortunate that this lab has many more datasets that they can share with us. Which I think will be good practice for ED puzzles.
[2:37 PM] beta_helix: (Sorry: vWa = von Willebrand factor type A)
[2:37 PM] HuubR:blood clotting?
[2:40 PM] beta_helix: These were the TNK1 puzzles (those proteins are implicated in the regulation of cell growth and cell death).
[2:41 PM] beta_helix: It will be very helpful to have a set of unsolved ED puzzles that we can run, especially when the new tools start being implemented.
[2:42 PM] Nicm25:i didn't know that we was challenging ED of thousands residues. we can it be posted as design puzzle? (and with known AA sequences) was it too early?
[2:42 PM] jmbrownlee333:Do many of them have crystallization domains, as the last one seemed to.
[2:43 PM] beta_helix: @Nicm25 we haven't been able to post any ED puzzles with more than 280 residues, so that was one main goal of this NSF grant. So if anyone knows any C++ experts who would be interested in improving Foldit Let us know! They don't have to be Foldit players
[2:44 PM] beta_helix: @jmbrownlee333 yes indeed
[2:44 PM] beta_helix: How difficult have the recent ED puzzles been?
[2:45 PM] beta_helix: Has the "radius" slider been useful, for example?
[2:45 PM] equilibria: I would say they have been hard, mostly because I could not tell where I was, as in the depth perception was terrible because of the monochrome background
[2:45 PM] jeff101:The sweet spot for the "radius" slider seemed too small. It would be good to have better resolution there.
[2:46 PM] equilibria:Also if i could chop up the electron density data so I could look at one section that would be great
[2:46 PM] jmbrownlee333:The last one was difficult for me because the dinsity seemed uneven. Some areas maddeningly bad. Other people seemed to deal with it better than me.
[2:46 PM] equilibria:^
[2:46 PM] jeff101:It moved from showing too little of the density to showing too much of the density.
[2:46 PM] HuubR:I think the big one (1964) was the third ED puzzle I played, so I was glad we had 2 weeks. Yes, I finally figured out how to make good use of the Radius slider.
[2:48 PM] jmbrownlee333:Better than no slider for sure!
[2:49 PM] beta_helix: Thank you, this is very helpful. It seems like the granularity on the slider could be improved. @equilibria being able to manipulate the ED cloud better is a major goal of the grant.
[2:49 PM] Susume (she/her): in the donut virus, radius took away the flat sides I was using to stay oriented, so I never used it. for others I used it sometiems
[2:49 PM] HuubR:The main point with this Radius was actually figuring out how the pivot point for the "camera" actually works. Up to then I used a sort of trial and error, but with 1964 I think I figured out how to use it properly.
[2:49 PM] jeff101:It would be nice if the Center/Align Protein on Density button would treat the protein like a rigid body and rotate and translate it to get a good fit to the ED/EM.
[2:50 PM] equilibria:I still havent figured out the radius works
[2:50 PM] beta_helix: I know many people have requested the ability to manually erase density (like the eraser tool in paint, but in 3D)(edited)
[2:51 PM] HuubR:@Susume (she/her): I made "flat sides" by folding the (reset) protein into an x, z, and y axis (x along the bottom, y along the top)
[2:51 PM] jeff101:It just has to vary 3 euler angles for rotation and 3 x y z variables for translation. The Nelder Meade SImplex Direct Search method could probably do it.
[2:51 PM] equilibria:wait i understood exactly 0% of that
[2:51 PM] equilibria:can you explain please?
[2:52 PM] jeff101:but you'd have to code it as part of FOldit, not as a recipe, because recipes don't have access to the ED vs x,y,z coordinates data.
[2:53 PM] jmbrownlee333:Wait, there's math involved in this. Wink. Wonderfully wonky Jeff101.
[2:54 PM] BletchleyPark: Would be useful to programmatically alter the rotation axis for a given residu
[2:54 PM] Susume (she/her): a little xyz graphic in the corner like you see in blender or other viewers would make keeping track of up/down etc easier
[2:54 PM] HuubR:https://fold.it/portal/files/chatimg/irc_977284_1615135836.png

[2:55 PM] beta_helix: There used to be a program called ProteinShop when I was in grad school, and it had 3 dials that you could manually move to rotate in the x/y/z planes.
[2:55 PM] Susume (she/her): huubr i see it's square, but it won;t be there when I am folding…?
[2:55 PM] sethcooper: i think it would be pretty straightforward to add axes
[2:56 PM] beta_helix: …but only for 64-bit windows
[2:56 PM] Susume (she/her): mouse is fine for rotating, what I want is a visual reminder of where up is
[2:56 PM] HuubR:Correct, I only used it while annotating the cloud
[2:56 PM] beta_helix: Got it, just the little arrows.
[2:57 PM] Susume (she/her): more dot colors would be nice too, I have started using double red, red-green, red-purple pairs etc to mean different things
[2:57 PM] HuubR::-)
[2:57 PM] beta_helix: That should be simple to add as well!
[2:57 PM] jeff101:how about orthographic instead of perspective. sometimes I want to align things from front to back, but perspective makes far away things appear closer on the screen than nearby things. Having orthographic would let far away things and nearby things be displayed with the same scale.
[2:57 PM] jmbrownlee333:I hate having one color be the default.
[2:58 PM] jmbrownlee333:for dots
[2:58 PM] HuubR:Oh yes: a setting "use last colour again"
[2:58 PM] jmbrownlee333:: )
[2:59 PM] sethcooper: i think there is an orthographic view already
[2:59 PM] sethcooper: it might be in the advanced view options
[2:59 PM] jeff101:please show me where the orthographic view is
[2:59 PM] beta_helix:Are there any last questions? Thank you all for coming!
[3:00 PM] jeff101:thank you for coming
[3:00 PM] equilibria:thanks for hosting!
[3:00 PM] Susume (she/her): thank you!
[3:01 PM] jmbrownlee333:I just want to upvote Jeff's complaint. "Center on density is worthless as is.
[3:01 PM] HuubR:Thank you, @sethcooper and @beta_helix, for an interesting hour!
[3:01 PM] beta_helix: @jmbrownlee333 thanks… I will double-vote that one
[3:01 PM] jmbrownlee333:Ya thanx guys
[3:01 PM] jeff101:Somewhere I wrote a feedback about how to use Nelder-Meade to fix the center on density tool.
[3:02 PM] BletchleyPark: rhank you
[3:02 PM] equilibria:nelder meade?
[3:02 PM] BletchleyPark: afk
[3:02 PM] equilibria:is that like a separate program?
[3:02 PM] jeff101:In matlab, it is implemented as the function fminsearch. It optimizes without using derivatives.
[3:03 PM] sethcooper: yes thanks everyone!
[3:03 PM] equilibria:what's matlab?
[3:03 PM] beta_helix: Thanks jeff101… I hope we'll hire someone soon to work on these suggestions!
[3:03 PM] sethcooper: i'm looking for the orthographic view option…
[3:03 PM] beta_helix: MATLAB is a program
[3:04 PM] HuubR:https://mathworks.com/
MathWorks - Makers of MATLAB and Simulink
MathWorks develops, sells, and supports MATLAB and Simulink products.
[3:04 PM] jeff101:Nelder and Meade apparently popularized it. It uses simplexes (triangles in 2-dimensions, tetrahedrons in 3-dimensions) and moves their vertices to find the best value of a function in multiple dimensions.
[3:04 PM] jeff101:THe dimensions could be angles, distances, band-strengths, etc.
[3:04 PM] beta_helix: FYI: Most schools have a free version of MATLAB.
[3:05 PM] beta_helix: Thanks again everyone… keep up the great folding!

[agcohn821]

[3:00 PM] horowsah: Hi all- I’m here for my office hour. If you don’t remember, I’m a general biochemist (mostly experimental, actually), and I’m mostly involved in ED and educational puzzles with Foldit, but I’m also good for answering general biochemistry questions. I’m not a protein design specialist, though.
[3:00 PM] donuts554:hello
[3:00 PM] horowsah: most of my work is on how nucleic acids affect protein folding, so that covers a lot of biochem
[3:02 PM] jeff101: thanks for coming
[3:02 PM] horowsah: no prob
[3:03 PM] Susume (she/her): I'll pose a non-foldit question then, can you tell ions apart in ED data? there are two xray structs in the PDB of the same protein with an ion visible in the ED, one team identified it as Cl-, the other identified it as Ca2+ (both were present)
[3:03 PM] horowsah: great question
[3:03 PM] jmbrownlee333:nucleic acids affect protein folding, and somehow i think of it the other way, that proteins affect nucleic acid folding. two -way street i guess
[3:03 PM] Susume (she/her): the sidechains were threonine, so I vote for Cl- (ca2+ likes oxygens)
[3:03 PM] horowsah: susume- i published a paper on that recently: https://journals.iucr.org/d/issues/2019/12/00/wa5123/
Acta Crystallographica Section D
Identifying dynamic, partially occupied residues using anomalous sc…
Structural studies of partially occupied, heterogeneous protein systems using crystallography are difficult. Here, methods are presented for detecting these states in crystals.

[3:04 PM] horowsah: short answer- it depends on the ion and the wavelength you use for crystallography. It's really easy to screw it up if you don't do the experiment just right (I've definitely done it before).
[3:05 PM] horowsah: jmbronlee: it is a two-way street. most people think of it the other way around than i do
[3:06 PM] alcor29:Is the massive presence of water fully accounted for or compensated for in any method of analysis or prediction?
[3:06 PM] Susume (she/her): I'll look at paper, ty
[3:07 PM] horowsah: alcor: water is a bugaboo of computational methods for sure. there are lots of models for it that people use, and none are perfect. it's one of the bigger limitations in computational biochemistry. but, it's still improving, and we can do a lot of things with the water models we have now
[3:07 PM] horowsah: in foldit, we use an implicit water model, which works fine for some things and is much faster than an explicit model where you would see all the individual water molecules surrounding the protein (would be awful hard to look at that way too)
[3:09 PM] donuts554:Can orange hydrophobic amino acids be exposed in proteins in water?
[3:10 PM] horowsah: so the dogma is that hydrophobics are buried away from water, but most proteins in nature aren't totally perfect, and there can be a stray hydrophobic here and there that is left on the surface. sometimes that will help in ligand binding
[3:11 PM] jmbrownlee333:Can u give any insight into the quality of ED data we are given . Sometimes they seem to be really crude (crappy phases? ) and other times better. I have similar questions around cryoEM but i don't understand those experiments as well. And what are the hopes for foldit players in the future.
[3:13 PM] horowsah: sure- ED data varies greatly in real life. sometimes it's awesome and high resolution and things are easier, and sometimes it's not. we still need to figure out how to fit it, even if it's not great data. however, that leads to a higher chance of it getting screwed up, which definitely happens. for crystallography, the quality of the data is mostly dictated by how perfectly the protein crystallizes. usually there are imperfections in the crystal that leads to them diffracting worse and giving worse data
[3:14 PM] horowsah: for cryo-em, it's a different set of factors, since you're not crystallizing a protein, but it also has to do with imperfections in the sample usually. for cryo-em, the technology is improving so rapidly that what was considered a great set of data 10 years ago is now thought of by crappy by today's standards.
[3:14 PM] jeff101: are there empty spaces inside the crystal between individual proteins?
[3:15 PM] horowsah: so there are spots where the proteins touch to form the crystal (called crystal contacts), and then there are open spaces as well that are filled with water inside the crystal, and are referred to as water channels
[3:15 PM] jmbrownlee333:Ya but somrtimes I think its the phasing not the diffraction holding us back.
[3:16 PM] horowsah: you can actually use those water channels to soak in things that bind to the protein
[3:17 PM] horowsah: for the phasing- yes, that's still an issue, but less often than it used to be, and as structure prediction improves, it will become less of a problem. That paper I sent in response to Susume's question earlier actually describes a crystallography beamline designed to phase with sulfur in cysteines, which will make it much easier to phase harder crystals.
[3:17 PM] jeff101: do the waters inside the crystal move around? do they show up in the ED data?
[3:18 PM] HuubR: To "soak in things" you would need pretty small molecules, not much bigger than water, right?
[3:19 PM] horowsah: the water does move if you collect the data above freezing, which sometimes is done, but sometimes we freeze crystals to help with diffraction (movement is tough). but we don't see the bulk water that's moving, only water that's in the same spot all the way through the crystal because it's bound to protein.
[3:19 PM] jmbrownlee333:Nice. we don't need no stinking selenoMet.
[3:19 PM] horowsah: For soaking, usually it's small stuff, but occasionally people will soak in things as big as whole peptides
[3:20 PM] alcor29:Can you make a comparison or analogy to 4K vs 8K in visual resolutions?
[3:21 PM] horowsah: hmm, i think so, but we almost never get to 4K quality.
[3:21 PM] horowsah: I'd say usually we're in VHS range
[3:22 PM] horowsah: …or worse, not sure what that would be
[3:22 PM] alcor29: What limits it from becoming better?
[3:23 PM] horowsah: sometimes it's inherent to the crystal form, sometimes it's limited by extra stuff getting stuck in the crystal. usually, crystallographers don't try to perfect the crystals, just get them to "good enough" which enables a model to be built at all. getting to perfection on the crystal could take decades or would never happen.
[3:24 PM] donuts554:Can Cryogenic Electron Microscopy be used to find the atomic structure of radioactive minerals?
[3:25 PM] alcor29: Going back to jmbs earlier question: " And what are the hopes for foldit players in the future."
[3:25 PM] horowsah: hmm, not positive on that one, and probably not yet; usually minerals are handled by crystallography since they usually are crystalline in pattern when pure
[3:25 PM] horowsah: ah- hopes
[3:26 PM] horowsah: well, we should be starting to introduce new ED features over the next few years that could vastly improve how much control Foldit players have in the process
[3:26 PM] horowsah: we'll want to see how much they help
[3:27 PM] jeff101: Foldit was awarded an NSF grant recently to improve its ED/EM abilities. The grant includes money to hire a programmer and encourages applicants who live near Seattle, Denver, or Boston. Will you be working with the programmer to improve Foldit?
[3:27 PM] horowsah: one of those features in particular will hopefully let foldit players improve the ED as part of the process. I should say this is a long way off.
[3:27 PM] horowsah: Yep, this is one of the tasks that the new programmer will be working on
[3:33 PM] jeff101: What can we expect in upcoming Foldit puzzles?
[3:34 PM] horowsah: hmm, good question. I actually can't mostly remember. I've been thinking mostly about the recent run of ED puzzles, and we likely will continue to have a few more coming up. Not sure on the protein design side.
[3:35 PM] jeff101: Can you tell how we've been doing on recent puzzles? How is our work helping science?
[3:36 PM] horowsah: I can give a pseudo update on the ED puzzles. We'll probably put together a blog post on this at some point, but not sure when. The short answer is, y'all have done really well on those. In some of these cases, it's outdoing scientists' work, in other cases, taking tough ED maps and managing to make them work when the scientists was having a tough time finding their way into it.
[3:36 PM] jmbrownlee333:Why are foldit players expected to be a good contribution to structure solving, compared to professionals and software and AI? I mean this question with respect, biut it feels like reinvention of the wheel, in some respects.
[3:38 PM] horowsah: For structure solving- software still is pretty bad, and nobody's made an AI to do it yet (not sure if and/or when that might happen and how it will do). The basic skills for putting stuff in ED that scientists develop is the same skill Foldit players develop. Except that a lot of scientists aren't very good at it. I'm actually maximally so-so at it. It's because I have terrible visual-spatial reasoning skills. So despite my good scientific thought processes, putting proteins in electron density is something I find very hard, and have spent years becoming merely passable at it as a result.
[3:41 PM] jmbrownlee333:So its a "more hands on deck" kinda thing?
[3:41 PM] jeff101: like 100 heads are better than one?
[3:42 PM] horowsah: in part, but it's also that Foldit players are likely just flat out better at it than a lot of scientists due to good visual spatial skills
[3:43 PM] jmbrownlee333:But sometimes I wish i could see what autobuild would do with the density.
[3:43 PM] horowsah: ah- so in the two ED papers we've published, we do compare to autobuild. Of all of the things we've tried, one kinda worked ok
[3:43 PM] jmbrownlee333:Ok, I like that answer. some people around here have awesome spatial recognition
[3:43 PM] jeff101: is it like in that movie "SLing Blade" where several guys are puzzled why a lawn mower isn't working. Then Billy Bob Thorton's character comes out, looks at it, and says "there's no gas in it" ?
[3:44 PM] horowsah: i've actually never seen sling blade, always wanted to
[3:44 PM] jmbrownlee333:touche
[3:44 PM] Susume (she/her): slingfold
[3:46 PM] jeff101: I wonder how Temple Grandin would do at Foldit.
[3:47 PM] horowsah: someday, it would be fun to create some fictional Foldit accounts of long-dead people we'd like to project how they'd be at Foldit. Like Picasso. I really want to see what proteins he'd design.
[3:47 PM] donuts554:How well can proteins conduct electricity?
[3:48 PM] horowsah: proteins are usually semi-conductive. DNA is actually a better conductor, if I recall correctly. I think that's actually a strategy used to find damage in DNA- shoot an electron down it and find where it stops, and that's the damage site
[3:50 PM] HuubR:Interesting. How do we know where the electron has stopped, can you see a space charge in an electron microscope?
[3:51 PM] horowsah: this has been a few years since i heard about it, but it's a protein that can shoot the electron down, and then another protein recognizes an extra electron elsewhere. That's assuming I'm remembering this correctly. It's a natural dna repair system.
[3:52 PM] HuubR:Amazing! :-o
[3:52 PM] donuts554:Thats somewhat interesting
[3:54 PM] pc:Hi. do we have some news in linker puzzle lab tests ^^
[3:54 PM] alcor29: Can shorter wavelengths be tried? short wave lazers?
[3:54 PM] alcor29:*lasers
[3:54 PM] horowsah: pc- sorry, i don't have any updates there, but i can pass along the request.
[3:55 PM] horowsah: alcor- can you clarify that a bit more?
[3:55 PM] pc: thanks ^^
[3:55 PM] alcor29: Don't know what you're using to get the densities. Assuming like shooting a target with x-rays.
[3:56 PM] donuts554:What things would you encourage Foldit players to do, and what things would you discourage Foldit players to do?
[3:56 PM] horowsah: ah yes- changing the wavelength helps for identifying things like ions and sometimes for phasing (not always), but it won't help with resolution usually
[3:57 PM] horowsah: donuts- probably others would answer the same thing: be creative and think outside the box. don't do the same thing as always, and don't just find one thing and keep drilling on it forever. try new things.
[3:57 PM] Susume (she/her): do you ever shoot the same sample with a series of different wavelengths to "see" different things?
[3:57 PM] alcor29: ditto
[3:57 PM] HuubR:What sort of wavelength is being used, is that X-ray, or gamma, …?
[3:58 PM] horowsah: susme- yep, we do that. the crystal will degrade, so we try not to do it too much
[3:58 PM] horowsah: so there's x-ray crystallography, and then there's also neutron crystallography where we actually shoot the crystals with neutrons
[3:58 PM] horowsah: turns out you can see hydrogens when you do that, whereas in x-ray crystallography you almost never can
[4:00 PM] Susume (she/her): in foldit densities we don't see the hydrogens but we can sometimes see density for some of the Hbonds, which is cool
[4:01 PM] horowsah: yep, in better ED that does happen
[4:01 PM] horowsah: ok, it's time for me to jet. thank you all for the great questions!
[4:01 PM] Susume (she/her): thanks for the free master class
[4:01 PM] alcor29: Thanks horowsah. Appreciated.

[agcohn821]

[1:02 PM] dudit: (he/him) When is the Office Hour started?
[1:02 PM] jeff101:Josh, what will you be studying?
[1:03 PM] sciren(he/him): Opens office door
[1:03 PM] sciren(he/him): How about now
[1:03 PM] jeff101: @Sciren: do you have any news for us? Like what aspects of FOldit are you working on?
[1:03 PM] sciren(he/him): By office door I mean open area table where I can get some sunshine, but same thing.
[1:07 PM] sciren(he/him): Just to give a little info, I as well as our team, are working hard to update the Small Molecule Builder so that we can get going with some small molecule puzzles. Which I very much hope you will all enjoy. Feel free to ask me any questions, and I will try to answer them.
[1:09 PM] sciren:(he/him) For those of you who may not have seen the design we are working towards, I will share it:
[1:09 PM] jeff101: do you mean like Puzzle 1855 Ligand Design ?
[1:10 PM] dudit (he/him): @Sciren I think the UI/UX needs a 'Mini Objectives' hover window, from @Formula350 idea design
[1:11 PM] sciren(he/him): Jeff101 similar to that one, but where that one uses the reaction design(which we are still planning on using in the future) This builder works in a more atomistic approach with the option of fragment addition.
[1:13 PM] sciren(he/him): Good point @Dudit there are some general UI changes in the works, and the Objectives panel is one of them that will be getting a bit of a makeover.
[1:16 PM] HuubR: I love the idea of the Objectives panel. Will it also be possible to move it into the corner, top left or top right?
[1:18 PM] sciren(he/him): Currently, one of the issues with the Objectives panel is that, in it's expanded form, it obscures the structure you are working on. So to help with that, a movement of the panel to the left-hand side of the screen in development.
[1:19 PM] jeff101:being able to move it would help. It is often in the way if I want to make a screenshot to post on the wiki.
[1:19 PM] jeff101:if you can move it, you could show the same info with less screenshots
[1:21 PM] jeff101:will you be adding LUA commands to tell atom types within a segment # (the segment could be a ligand molecule or an amino acid) ?
[1:21 PM] sciren(he/him): Indeed Jeff101, and we are looking to simplify the menu to show only the directly pertinent information(where scoring is involved), with the ability to get more information should you want it.
[1:22 PM] jeff101: GetAtomType(segnum,atomnum). Have it return C N O S H for amino acid atoms, for example.
[1:24 PM] jeff101: for ligands, perhaps do more complicated atom names, like you might find in a pdb file. They could help distinguish carbons in C=O from carbons with 4 single bonds, for example.
[1:25 PM] sciren (he/him): Currently that isn't in development, but I can definitely take a note of it, and discuss it with the team to see what can/should be done.
[1:25 PM] sciren (he/him): Thank you for the suggestion!
[1:26 PM] jeff101:or be able to query what atoms each atom in a ligand is bonded to
[1:27 PM] HuubR: Would this GetAtomType(segnum,atomnum) also account for different numbers of atoms, e.g. for the first and last segments, or with disulfide bonds?
[1:28 PM] jeff101:and for each atom listed, it could list the type of bond used to connect the atoms: single, double, triple bond, aromatic ring bond (somewhere between single and double bond), etc. maybe it could be a # like 1 of single bonds, 2 for double bonds, 3 for triple bonds, and non-integers for bonds in between
[1:30 PM] jeff101:maybe hydrogen bonds and disulfide bonds would have their own types
[1:32 PM] sciren(he/him): For scripting purposes, I can definitely see why the "more information the better" We will have to look at what would be needed to provide this. I really like the idea of being able to get/share this information on an atom by atom basis.
[1:36 PM] jeff101:it would be nice to be able to measure the distances between atoms on the same ligand. also would be good if we could add bands between atoms on the same ligand.
[1:38 PM] sciren(he/him): Good call Jeff101! These were actually a couple of features that have been discussed based on preliminary testing and both of these items are actually being looked into.(edited)
[1:41 PM] jeff101: that's good
[1:43 PM] jeff101: are there good rosetta forcefields for general small molecules and ligands?
[1:45 PM] jeff101: can they handle metals (copper, iron, zinc), calcium, sodium, or free radicals ?
[1:49 PM] sciren(he/him): That is a good question jeff101. Unfortunately I am not the best versed in the Rosetta force field and its limitations. I know that the force field that is being used better that previous iterations for small molecules. I will looks into it and let you know what I find.(edited)
[1:52 PM] jeff101:are you at Vanderbilt? what is your training?
[1:56 PM] sciren(he/him): I am indeed! So a little about my background: I spent some time outside of the academic world(mostly in the service industry) and eventually decided to go back to school and finish my chemistry degree. Along the way I realized I was very interested in computer science. So I thought why not just do both? I engaged in undergraduate research that was pairing gaming technology, and molecular modeling software, and loved the work I was doing. So when I graduated, I threw a long shot search into google for jobs involving video games and chemistry, and stumbled upon a job listing to work on Foldit.
[1:58 PM] sciren (he/him): Now I get to combine two of my favorite things: Gaming and science.
[1:59 PM] jeff101: that's good
[2:01 PM] sciren(he/him): What's also good is all of the amazing solutions to very challenging problems that you all come up with. It is an absolute joy to see what you all come up with.
[2:01 PM] jeff101:when do you think the next reaction design / small molecule design puzzle will be ?
[2:02 PM] sciren (he/him): We're hoping to get it out to you all as soon as we can, but there are some items to take care of before then.
[2:02 PM] Josh jeff101 sorry, I was in a meeting. Just got your question on what I'm studying. I'll be overhauling the tutorial based on the latest and greatest research in instructional design theories, so I'll need to ask some veterans to walk me through their strategies so I can teach how the pros play.
[2:03 PM] dudit(he/him):( How about more COVID puzzles?
[2:04 PM] jeff101:sounds good, Josh
[2:05 PM] Nicm25: Oh, was that office hours?
[2:06 PM] jeff101:yes Nicm25
[2:06 PM] jeff101: got any questions for Josh or sciren:?
[2:06 PM] sciren(he/him): @Dudit I'm not sure about any COVID puzzle on the horizon, but that can be subject to change.
[2:07 PM] sciren(he/him): Yes indeed. I can hang around for another couple more if you have any Nicm25
[2:08 PM] Nicm25:I just joined IRC and missed conversation…lol
[2:08 PM] jeff101: sciren is working small molecule design tools for Foldit. Josh is working on Foldit Tutorials and teaching new players how to play.
[2:10 PM] dudit (he/him): @Sciren small molecule design COVID puzzle?
[2:12 PM] sciren(he/him): @Dudit currently not that I know of, but that can be subject to change.
[2:13 PM] sciren(he/him): I do want to thank everyone for the questions and suggestions. I'm going to leave my sunlit office for one with less windows, and more coding. I hope you all have a great one, and happy folding!
[2:13 PM] jeff101:thanks for coming. Have a good day.
[2:13 PM] dudit (he/him): Thank you so much @Sciren

[agcohn821]

[10:59 AM] bkoep: Hello all, and welcome to Foldit Office Hours!
[11:00 AM] Dudit (he/him): Hi @bkoep
[11:00 AM] cjddig (he/him): Hello bkoep!
[11:00 AM] alcor29:Good morning @bkoep. 1. What is the correlation between confidence and similarity? What exactly is the difference between them? Why in your blog fig.3 did you choose to test with confidence and not similarity? Is confidence based on the similarity number? If so why not just use similarity? 2. Can you respond to Boots' comments on the AF blog, please? https://fold.it/portal/node/2011929
The AlphaFold prediction tool in Foldit
The AlphaFold prediction tool in Foldit
[11:00 AM] bkoep: How are people finding the new AlphaFold tool?
[11:00 AM] donuts554: hello
[11:00 AM] alcor29: Murky.
[11:02 AM] spvincent: hi bkoep
[11:03 AM] donuts554: I havent tried the new AlphaFold tool yet
[11:03 AM] spvincent: It seems to work very well, assuming the predictions are accurate. Apart from some solutuons crashimng when you try to load them
@alcor29 Good morning @bkoep. 1. What is the correlation between confidence and similarity? What exactly is the difference between them? Why in your blog fig.3 did you choose to test with confidence and not similarity? Is confidence based on the similarity number? If so why not just use similarity? 2. Can you respond to Boots' comments on the AF blog, please? https://fold.it/portal/node/2011929
[11:03 AM] bkoep: 1. Confidence and similarity are not necessarily related. Confidence is unrelated to the structure of your solution, and it depends on your sequence only. Similarity does depend on the structure of your solution. So, you could take a well-folded sequence and paste it onto an extended chain, and it would have high confidence but low similarity. If you folded up that sequence and ran it again, the second run would give you the same confidence (because the seq is the same) but the similarity should be higher (if it is folded correctly)(edited)
[11:04 AM] Dudit (he/him): @bkoep Is it better to further developing RoseTTAFold to be better at Protein Design puzzle accuracy?
[11:05 AM] bkoep: … Benchmarking AF by the similarity metric is not so informative, since there is an easy way to game the similarity metric without improving your design (just load an AF prediction, submit it again, and you will get very high similarity)
[11:06 AM] cjddig (he/him): 1. Will AlphaFold work if binder is too far from target? Can it automatically make binder closer to target? 2. Does AlphaFold consider Objectives? 3. Why does AlphaFold solution have some clashes? Doesn't it optimize the protein?
[11:06 AM] BletchleyPark: To what would this be similar ?
[11:07 AM] spvincent: Is there a need anymore for some of the filters in design puzzles? Ideal Loops, Core, etc. could all be handled by AF
[11:07 AM] Dudit (he/him): @bkoep Is it possible that a protein has a 100% confidence & 100% similarity?
[11:08 AM] bkoep: Ah yes, I hadn't seen Boots' comment yet. The short answer is that AF is probably just highlighting some Foldit quirks. I will write out a fuller response after the office hour.
[11:09 AM] alcor29: When you say structure do you mean just the helix, strand or the actual tertiary fold? How does it measure this. ?
[11:09 AM] Dudit (he/him): @bkoep What's the difference between AlphaFold2 vs RoseTTAFold?
[11:10 AM] alcor29: Same thought as spvincent. Use AF to handle loops etc.
[11:10 AM] BletchleyPark: @bkoep Is this correct: AF bases its design on the sequence only. While designing a backbone, changing the sequence through mutating will change the AF prediction and thus the backbone.
@Dudit (he/him)
@bkoep Is it better to further developing RoseTTAFold to be better at Protein Design puzzle accuracy?
[11:11 AM] bkoep: There are machine-learning (ML) scientists always working to improve RoseTTAFold and new algorithms for protein design! AlphaFold2 and RoseTTAFold are both neural networks, but their precise architecture is very different. Unfortunately, I don't have enough background in ML to explain the full differences between the two. I think there are some informative blogs out there that may help you understand the nuances of these networks, for example Mohammed AlQuraishi's blog here: https://moalquraishi.wordpress.com/2021/07/25/the-alphafold2-method-paper-a-fount-of-good-ideas/
Some Thoughts on a Mysterious Universe
Mohammed AlQuraishi
The AlphaFold2 Method Paper: A Fount of Good Ideas
Just over a week ago the long-awaited AlphaFold2 (AF2) method paper and associated code finally came out, putting to rest questions that I and many others raised about public disclosure of AF2. Alr…
@cjddig (he/him)

1. Will AlphaFold work if binder is too far from target? Can it automatically make binder closer to target? 2. Does AlphaFold consider Objectives? 3. Why does AlphaFold solution have some clashes? Doesn't it optimize the protein?
   [11:13 AM] bkoep: 1. Right now, it doesn't matter if the binder is far from the target in your Foldit solution. We are only passing the binder sequence to AlphaFold, so it doesn't know anything at all about the target. There are other researchers developing new ways to use AF with binder design (i.e. ways that do include the target) so that AF can predict not only binder folding, but also the binding of the two separate chains.
   @cjddig (he/him)
2. Will AlphaFold work if binder is too far from target? Can it automatically make binder closer to target? 2. Does AlphaFold consider Objectives? 3. Why does AlphaFold solution have some clashes? Doesn't it optimize the protein?
   [11:15 AM] bkoep: 2. No AF does not consider Foldit Objectives in any way. When you upload a solution for AlphaFold, we simply extract the sequence of your solution and pass that as input to the AF algorithm. Then the AF algorithm spits out a predicted structure, and we send that back to your Foldit client.
   [11:16 AM] alcor29: So I get why AF gets a confidence number. Still not clear about why similarity matters then?
   @cjddig (he/him)
3. Will AlphaFold work if binder is too far from target? Can it automatically make binder closer to target? 2. Does AlphaFold consider Objectives? 3. Why does AlphaFold solution have some clashes? Doesn't it optimize the protein?
   [11:17 AM] bkoep: 3. No, we are only running the raw AlphaFold neural network, and we leave the optimization to you! The AlphaFold team recommends some form of optimization on raw predictions (in CASP they used the Amber molecular dynamics force field). So the clashes are to be expected.
   [11:18 AM] Dudit (he/him): @bkoep Should we aim to design a protein that has a 100% in both of confidence & similarity?
   [11:19 AM] alcor29:Also, I mutate many times during a design after some things change. But I see you prefer just sending after the final mutate, but, the AF return may cause me to mutate again. Is this a problem for you?
   spvincent: Is there a need anymore for some of the filters in design puzzles? Ideal Loops, Core, etc. could all be handled by AF
   [11:20 AM] bkoep: This is an interesting question (Do we still need filters in design puzzles?)… Maybe not, if the AF predictions are accurate enough and Foldit players are able to work with AF. Currently, my impression is that the AF outputs are pretty opaque, and don't give much feedback about how a problematic design might be improved. We have some ideas about this, but for now I think the Objectives are still effective guidelines to help new players come up with reasonable designs.
   @Dudit (he/him)
   @bkoep Is it possible that a protein has a 100% confidence & 100% similarity?
   [11:22 AM] bkoep: No, I don't think you would ever see an AF prediction with 100% confidence (although I'm not certain whether this is theoretically possible with the network). The highest I've seen for a designed protein is about 95% confidence.
   [11:22 AM] Vincera:I had a 96.8 on one transition.
   [11:22 AM] alcor29: Just a thought: Can the aa mutate load for each puzzle be loaded with AFs high confidence loop structures?
   @alcor29
   When you say structure do you mean just the helix, strand or the actual tertiary fold? How does it measure this. ?
   [11:24 AM] bkoep: Sorry, I just saw this followup… AF similarity is measured against the entire protein fold (i.e. tertiary structure). In this case we are using a similarity metric called lDDT (local distance difference test. I think?) that computes the internal distance between all (i,j) residue pairs and compares that against the internal distances in the reference structure
   [11:25 AM] BletchleyPark: @bkoep, how many GPUs are ussed to infer the structure for a given sequence ?
   [11:25 AM] Vincera:Have you viewed the AF submissions from yesterday (eg, my 96,8 transition)?
   [11:25 AM] BletchleyPark: and thank you for the similarity answer
   @bkoep Is this correct: AF bases its design on the sequence only. While designing a backbone, changing the sequence through mutating will change the AF prediction and thus the backbone.
   [11:26 AM] bkoep: Yes! The output prediction from AF depends on sequence only. If you fold up the same sequence in multiple ways, you will get the same prediction and confidence every time (but similarity will differ). As soon as you make one mutation, you can expect a different prediction from AF.
   [11:27 AM] BletchleyPark: thank you ! @bkoep, how long did it take you to train the network (on how many GPUs)and did you use the entire PDB for this ?
   @alcor29, So I get why AF gets a confidence number. Still not clear about why similarity matters then?
   [11:30 AM] bkoep: Similarity is most meaningful if you have designed your protein completely "blind" to the AF prediction (i.e. 100% in Foldit, without loading AF predictions). Since Foldit and AF are completely independent, this would mean your design is very promising if both Foldit and AF agree (it has a high Foldit score and a high AF confidence). However, if you load in an AF prediction and continue working off of that, future AF similarity metrics will not be as meaningful, since you have now biased your design to be similar to the AF prediction.
   [11:31 AM] alcor29: Exactly. So why even bother with similarity figures?
   [11:31 AM] spvincent: I'm not sure if the output from AF is opaque. Just loading the solution and toggling back and forth between it and the source (Score/Hydro colouring helps here) gives a good indication of what the problematic areas are.
   @Dudit (he/him)
   @bkoep Should we aim to design a protein that has a 100% in both of confidence & similarity?
   [11:32 AM] bkoep: Most of our experiments so far suggest that 80% AF confidence is probably good enough for protein designs. If you reach 80% confidence, then I would recommend working on a new design!
   [11:32 AM] alcor29: Thanks @bkoep.
   [11:32 AM] cjddig (he/him): I previously tested AF with triple helices protein which has some threonine on its core, and AF returned a long helix. When we make a unique structure, such as beta helix, is there any possibility that AF would give us totally different structure like that?
   [11:32 AM] Dudit (he/him): Thank you @bkoep
   @alcor29, Also, I mutate many times during a design after some things change. But I see you prefer just sending after the final mutate, but, the AF return may cause me to mutate again. Is this a problem for you?
   [11:33 AM] bkoep: No, that sounds perfectly fine! This is the way we anticipated players would probably use AF (i.e. get an AF prediction, load it into Foldit and mutate/optimize, get another AF prediction, do more mutate/optimization, etc.)
   @alcor29, Just a thought: Can the aa mutate load for each puzzle be loaded with AFs high confidence loop structures?
   [11:35 AM] bkoep: @alcor29 I'm not sure I understand this question, could you rephrase it? What do you mean by "aa mutate load" and "high confidence loop structures"?
   BletchleyPark: @bkoep, how many GPUs are ussed to infer the structure for a given sequence ?
   [11:37 AM] bkoep: Right now, we are using a single GPU for inference. It might be an RTX2080 (?), I'm sorry I'm not certain about the hardware
   [11:37 AM] alcor29: You know how the table load tells it that GLY, ALA can't be used if the ss is a helix , but it knows that it's OK for a loop. So maybe it can be loaded to recognize this is a loop and only the AF high confidence aas in loops are allowed so it would grey out all the other aas? Like I said . JUst a thought. Might be too complicated to implement?
   [11:37 AM] BletchleyPark: sounds good
   [11:38 AM] BletchleyPark: sounds like we could run this in our own systems locally
   Vincera:Have you viewed the AF submissions from yesterday (eg, my 96,8 transition)?
   [11:38 AM] bkoep: @Vincera I have not yet taken a look at the most recent AF submissions. But 96.8% is a very impressive confidence! What kind of fold was this? A helical bundle or something else?
   [11:41 AM] Vincera:Quad, from what I recall. I submitted several diif…
   @cjddig (he/him) I previously tested AF with triple helices protein which has some threonine on its core, and AF returned a long helix. When we make a unique structure, such as beta helix, is there any possibility that AF would give us totally different structure like that?
   [11:41 AM] bkoep: @cjddig, do you mean you tested AF with a natural protein? Note that our "slimmed down" version of AlphaFold will not perform as well as the "official" AlphaFold algorithm that was used for CASP – especially for natural protein targets.
   [11:42 AM] cjddig (he/him): Ah, I designed the protein, not a natural one.
   @alcor29, You know how the table load tells it that GLY, ALA can't be used if the ss is a helix , but it knows that it's OK for a loop. So maybe it can be loaded to recognize this is a loop and only the AF high confidence aas in loops are allowed so it would grey out all the other aas? Like I said . JUst a thought. Might be too complicated to implement?
   [11:45 AM] bkoep: Ah, gotcha. Yes, that makes sense. Potentially, we can do even better… In principle, we should be able to tweak the AF network so that it returns AA suggestions at each position that would increase its confidence. Very similar to what was reported in the trRosetta energy landscape optimization paper, but this research is ongoing…
   @cjddig (he/him)
   Ah, I designed the protein, not a natural one.
   [11:47 AM] bkoep: We are still learning the limitations of AF, but so far it has been very accurate for small single-domain designs (like the type we are designing in Foldit). If AF predicted a completely different structure for your design, I'm afraid that probably means there are problems with your design. We definitely don't like THR in the protein core, especially for helical bundles…
   [11:47 AM] alcor29: Wow.
   [11:50 AM] alcor29: Can we have more time for 2029?
   [11:51 AM] donuts554: How does the Backbone subscore work in the Foldit scoring system?
   @alcor29 Can we have more time for 2029?
   [11:51 AM] bkoep: Would more time be useful for reaching designs with better AF confidence? Or is there another reason for more time?
   @bkoep
   Would more time be useful for reaching designs with better AF confidence? Or is there another reason for more time?
   [11:52 AM] alcor29: Only since it's new for us and we are just feeling our way into it.
   [11:53 AM] alcor29: For the future. I guess we will see how it goes.
   [11:53 AM] cjddig (he/him): My last question: do you think that AF will change the way of folding a lot?
   [11:53 AM] CrossedSticks (she/her): Personally I don't need more time, I prefer to tackle a new puzzle rather than keep rehashing the existing ones
   [11:54 AM] BletchleyPark: @bkoep, is there away to have the AF result be shown as ready upon completion. I now use reload repeatedly
   [11:54 AM] donuts554: I think the AlphaFold feature in Foldit is like a thermodynamic simulating feature in Foldit
   donuts554: How does the Backbone subscore work in the Foldit scoring system?
   [11:54 AM] bkoep: @donuts554, the backbone subscore evaluates the bond torsions of bonds along the protein backbone. This mostly reflects the (phi, psi) torsion preferences in the Rama Map, but also includes terms to make sure that the peptide bonds (i.e. the bonds between residues) are planar
   @cjddig (he/him) My last question: do you think that AF will change the way of folding a lot?
   [11:56 AM] bkoep: This is the big unknown! AF definitely promises to open up some new opportunities. For example, I am hoping AF gives us the confidence to branch out into more diverse protein folds (not just helical bundles and ferredoxin-like folds)
   [11:56 AM] alcor29: CrossedSticks has a point. I get tired of looking at the same puzzle after a bit.
   [11:57 AM] robgee: Will the results of the alphafold puzzles be fed back into the alphafold neural network ?
   BletchleyPark: @bkoep, is there away to have the AF result be shown as ready upon completion. I now use reload repeatedly
   [11:57 AM] bkoep:
   @BletchleyPark, currently there are no notifications for completed AF jobs. If the AlphaFold panel is open, then it will auto-refresh every 30s
   [11:57 AM]
   BletchleyPark: thank you, that is good to know already
   [11:58 AM] Nicm25:(I refrained from asking the question) okay, so if I should use AlphaFold2 on its own to get more detailed information about protein I'm designing, should I use it? as long as AlphaFold's server uses only neural networks, if it can't compensate for interpolated information such as sequence matching and optimization, were we going to get it done? Or was it premature to suggest it? of course, only in puzzles where there is AlphaFold button on foldit. (no need to answer.)(edited)
   robgee: Will the results of the alphafold puzzles be fed back into the alphafold neural network ?
   [12:01 PM] bkoep: @robgee, that's an interesting idea. The AF network is trained only on experimentally-solved protein structures. I would worry a little bit about including "unverified" protein structures in the training set. I'm not sure if there would be anything to gain from that
   [12:09 PM] BletchleyPark: @bkoep, I see a user called AlphaFold has this been a test account
   [12:09 PM] BletchleyPark: ?
   @Nicm25 (I refrained from asking the question) okay, so if I should use AlphaFold2 on its own to get more detailed information about protein I'm designing, should I use it? as long as AlphaFold's server uses only neural networks, if it can't compensate for interpolated information such as sequence matching and optimization, were we going to get it done? Or was it premature to suggest it? of course, only in puzzles where there is AlphaFold button on foldit. (no need to answer.)(edited)
   [12:10 PM] bkoep: Hmm, I'm not sure I follow… Are you suggesting using AlphaFold2 outside of Foldit to improve your design? Technically it is against the Foldit rules to copy a non-Foldit protein into the game, and we encourage players to do all of their work within Foldit (this is important for the integrity of Foldit as a competitive game). But I think you should use whatever tools you want to get ideas about how to improve your Foldit strategies. I see no problem if you want to do some exploring with AlphaFold2 outside of Foldit.
   BletchleyPark: @bkoep, I see a user called AlphaFold has this been a test account
   [12:11 PM] bkoep: Hmm, I'm not sure. It's not my test account, anyway
   [12:13 PM] Nicm25: bkoep, sorry for hassle, and thanks for answering. :-)
   [12:14 PM] cjddig (he/him): Thanks for answering bkoep!
   [12:14 PM] Dudit (he/him): Thank you so much for the Office Hour @bkoep
   [12:16 PM] bkoep: Yes, it looks like my hour is up
   [12:16 PM] bkoep: Thanks all for the great questions!
   [12:16 PM] BletchleyPark: we had bonus time, thank you !
   [12:16 PM] donuts554: You are welcome!
   [12:17 PM] bkoep: If you have any more questions, please don't hesitate to post comments on the website. Or you can ping me on Discord!

[agcohn821]

[3:02 PM] beta_helix: Hi everyone! Thank you for joining us today for Foldit Office Hours. Andreas just joined the Foldit team, and is working on improving Electron Density Tools (as you can see here, if you are on Discord
[3:03 PM] Dudit (he/him): Hello @beta_helix
[3:04 PM] beta_helix:
Has anyone had a chance to try the new "Select Cloud" devprev feature yet?
[3:05 PM] alcor29: Just tried. And it crashed.
beta_helix: Has anyone had a chance to try the new "Select Cloud" devprev feature yet?
[3:05 PM] Nicm25: I already tried it out 40 minutes ago, with brief review hanging out on this (veteran)channel.
[3:05 PM] beta_helix: Are you in the Selection Interface?
[3:07 PM] alcor29: Nope. Was in original. Will switch. Tx.
[3:07 PM] beta_helix: Whew! Yes, sorry… we should have warned about that (I'll add that to the newspost comments)
[3:08 PM] petridecus: yea we already in the process of removing the option from the original interface, whoops
[3:09 PM] beta_helix: Random poll: How many of you use the Original Interface (most of the time)?
[3:09 PM] spvincent: me
[3:09 PM] Dudit (he/him): me too
[3:10 PM] Nicm25: oops, I'm using the original interface to do protease (cut)
[3:10 PM] beta_helix: Were all of you able to use the Select Cloud feature by switching to the Selection Interface?
[3:11 PM] alcor29: about 50/50 or 60original/40selection
[3:12 PM] alcor29: yup
[3:13 PM] alcor29: I click on select cloud. AAs change color but that's it.
[3:14 PM] spvincent: guess so. not all that familiar with ED puzzles though.
[3:14 PM] beta_helix: @Nicm25 you already posted you thoughts on the tool. Is anything missing or do you have any suggestions on how it could be improved?
[3:15 PM] Nicm25: I refrained from speaking until the documentation was in perfect order, which is bit of surprise!?
[3:16 PM] Nicm25: today, for some time we have not had any density puzzles posted, so I don't think players will be able to answer about this.
[3:17 PM] beta_helix: We wanted all your thoughts before posting it to main, which is when we release the documentation.
[3:17 PM] beta_helix: You are right, I should have pointed out the Beginner ED Puzzle [3:17 PM] beta_helix: Sorry about that, I forget that everyone has that checkbox turned off!
[3:18 PM] petridecus: the beginner ED puzzle is perfect for testing out this tool, that's what i used while i was developing the feature as well
[3:19 PM] alcor29: Trying to follow the petridecus gif, but no luck. Missing an action somewhere.
[3:20 PM] Nicm25: ummm, I am still testing with 2007 and even older puzzles on how I can go beyond algorithm I came up with, solutions I solved manually. ;D
[3:21 PM] beta_helix: @alcor29 what OS are you on?
@alcor29: Trying to follow the petridecus gif, but no luck. Missing an action somewhere.
[3:21 PM] petridecus: what's the issue that you're running into. are the residues not being selected?
@Nicm25: ummm, I am still testing with 2007 and even older puzzles on how I can go beyond algorithm I came up with, solutions I solved manually. ;D
[3:21 PM] beta_helix: We hope to provide even more ED tools soon
[3:21 PM] alcor29: Win10
[3:22 PM] petridecus: yeah there's a lot more useful stuff that's already in the works!
[3:22 PM] alcor29: Cannot select pieces of the cloud as in the gif
[3:23 PM] petridecus: @alcor29 so you press the button and nothing happens then?
[3:24 PM] alcor29: Nope. the colors change but you seem to be selecting the area you want by moving the protein. I try that and nothing happens. Even though I also clicked center on density.
[3:24 PM] petridecus: the area of selection does not move until you press the button again
[3:25 PM] alcor29: No. If it works for everybody else we should go on. Eventually I'll figure it out.
[3:26 PM] beta_helix: Well, this is helpful… because we need to make sure it is clear and easy to use for everyone
[3:26 PM] alcor29: OOps just got it o work I think I need to double click.
[3:26 PM] beta_helix: Yay!
[3:27 PM] alcor29: OK. It highlitghts different sections but the cloud does not change so what is gained?
[3:28 PM] alcor29: Must still be missing something.
[3:28 PM] beta_helix: That's why were are here
[3:30 PM] alcor29: Do I need to change my radius and thresholds to see the effect?
[3:30 PM] petridecus: in short, it's using the cloud as a means to select and then focus your wiggle/shake operations on a particular region of the protein

[3:30 PM] petridecus: the goal is not to change the cloud itself or anything like that
[3:32 PM] petridecus: the cloud is a very visually informative element to the puzzles in which it's present- we believe that also being able to select regions using that visual element could make it even more useful
[3:33 PM] HuubR: Please correct me if I'm wrong; I'm not sure I understand this correctly. When using Selection Interface, the [Select Cloud] button will select all residues that are within the cloud, and deselect all that are outside?
[3:34 PM] petridecus: yup that's correct!
[3:34 PM] alcor29: Agree. But don't see any changes in the cloud so maybe 2007 shows no effect because it is already about as centered as it can be?
[3:35 PM] alcor29: Looking for different ed puzzle.
[3:35 PM] petridecus: again, the goal is not to change the cloud itself. not sure if i understand your question
[3:35 PM] alcor29: Beginner.
[3:36 PM] HuubR:But when I move the entire protein out of the cloud, and then press the button, it still selects everything (157 residues in the Beginner puzzle). What am I doing wrong? Wrong OS (Win 8)?
[3:36 PM] alcor29: OK, I don't have a beginner ED puzz on my list.
[3:37 PM] petridecus: @alcor29 make sure you've checked the box to show beginner puzzles
HuubR:But when I move the entire protein out of the cloud, and then press the button, it still selects everything (157 residues in the Beginner puzzle). What am I doing wrong? Wrong OS (Win 8)?
[3:37 PM] petridecus: that may be a bug! i'll try to reproduce it
[3:37 PM] alcor29: yup all checked beginner, expired etc.
[3:38 PM] alcor29: 2007 was the only ED on my menu
[3:39 PM] beta_helix: Staphylococcus Aureus Electron Density
[3:39 PM] beta_helix: strange, I thought those were available to all players
[3:40 PM] alcor29: Staph works. ED description ran off the page.
[3:40 PM] beta_helix: woo-hoo
[3:42 PM] beta_helix: @HuubR any other feedback or possible bugs?
[3:42 PM] alcor29: OK. Seems to working. Sorry.
[3:42 PM] HuubR:No, not yet. I just found the puzzle 15 minutes ago :-)
@alcor29 OK. Seems to working. Sorry.
[3:43 PM] petridecus: if there are issues that were specifically happening in the 2007 puzzle that would still be good to know. is the tool behaving differently between the two puzzles?
[3:45 PM] Nicm25: in my environment, it worked without any problems, as far as I understand.
@alcor29 OK. Seems to working. Sorry.
[3:45 PM] beta_helix: No worries at all. Thank you! We clearly want to post a detailed description of this before it goes to main (ie in a blog or on the wiki)
[3:47 PM] alcor29: @petridecus Can't really say. But I didn't see any effect that was new on 2007
[3:47 PM] alcor29: It will take a while before I actually understand this by fooling around with it a bit.
[3:48 PM] Nicm25: well, I would like to ask question about direction. do we want tools to change the pose of proteins based on their density? or do we want tools to make the density more visible or new ways to draw it?
[3:49 PM] beta_helix: @Nicm25 Great question! What do you think?
[3:51 PM] Nicm25: Ah, I wanted to ask about which is priority, I can give idea but I need to put it together as documents.
[3:52 PM] petridecus: current development, including this tool, are definitely leaning more towards the "changing the pose" direction
[3:52 PM] alcor29: Bottom line question? How is this expected to improve ED folding?
[3:53 PM] jeff101: how do all the copies of the ED that we can't see affect the results of the Select Cloud tool?
[3:53 PM] alcor29: A specific example.
[3:54 PM] petridecus: @alcor29 it's a tool that we hope will give players more precision when working on certain regions of the protein- when doing wiggle/shake operations in particular
[3:55 PM] Nicm25: (my concern was whether or not I could provide new 'fair tool'. so I am in process of sifting through ideas to pass on to new developers.)
[3:55 PM] petridecus: if someone is looking at a certain region of the protein and wants their operations to only affect that region, it's now as simple as moving the cloud over that area and pressing the select cloud button
[3:56 PM] beta_helix: especially when we give you giant electron density clouds!
[3:56 PM] jeff101 if we select region A of the visiblecloud, is it also selected in all the hidden clouds? then, if the protein is extended so it crosses multiple copies of region A of the ED cloud, will multiple chunks of the protein get selected too?
[3:57 PM] jeff101 not just the part of the protein in the visible copy of region A of the ED cloud?
[3:57 PM] petridecus: i don't understand this question. when you refer to hidden clouds, do you mean the different display options for the cloud?
[3:57 PM] jeff101: if you drag the protein thru seeminly empty space, you will find many regions wth nonzero ED
[3:57 PM] petridecus: i can say this for sure though- once the residues are selected based on the location of the cloud, the selection is no longer dependent on where the cloud is
[3:57 PM] beta_helix: @jeff101 the symmetric copies of the cloud should stay the same
[3:58 PM] beta_helix: but that is a good question about what remains selected…
[3:58 PM] beta_helix: we'll have to check that, but my guess is that only the "selected protein" actually gets selected
[3:58 PM] beta_helix: @jeff101 have you had a chance to try it out yet? If so, thoughts?
[3:59 PM] jeff101: the protein can straddle more than one ED cloud. part of the protein can be in the visible ED cloud. other parts of the protein can be in the hidden ED clouds.
[3:59 PM] jeff101: I have not tried the new ED tools
[3:59 PM] beta_helix: right… but the selected region, should only be what you see. Key word being "should"
[4:00 PM] beta_helix: Got it. Just curious, as I know you have many ED suggestions.
[4:00 PM] beta_helix: petridecus has your wish list
[4:00 PM] jeff101: that's good
[4:01 PM] alcor29: @petridecus when you say move the cloud over an area what do you mean. I can move the area to the cloud but not vice versa.
[4:01 PM] HuubR: jeff101, when you move the protein out of the ED cloud, you will see the score drop. Have you seen it go up again when the protein was not in the visible cloud?
[4:01 PM] beta_helix: Yes, HuubR, that is possible.
[4:01 PM] petridecus: when you move the focus of the camera over a certain region of the protein, thus causing that part of the cloud to be visible @alcor29
[4:01 PM] alcor29: Suggestion: I think a live demo would be most usefull
[4:02 PM] beta_helix: Great idea!
[4:02 PM] beta_helix: Andreas, you down for that before this goes to main?
[4:02 PM] jeff101: I have seen the score rise and fall and rise again when I drag the protein through space
[4:02 PM] petridecus: yeah absolutely
[4:02 PM] beta_helix: Sweet… sounds like a plan.
[4:03 PM] beta_helix: Thank you all for your feedback on this, and for stopping by today's Office Hours.
[4:03 PM] alcor29: Thank you all.
[4:03 PM] beta_helix: Unless there are any last questions…
[4:03 PM] jeff101: crystals are periodic structures. for simple ones like salt, there are unit cells
[4:03 PM] petridecus: thanks everyone!
[4:03 PM] Dudit (he/him): Thank you @beta_helix @petridecus
[4:03 PM] jeff101: the unit cells repeat in all directions
[4:03 PM] Nicm25: thank you very much.
[4:04 PM] jeff101: yes, thanks
[4:04 PM] HuubR:Yes, thank you all :-)
[4:04 PM] beta_helix: @jeff101 sometimes what we do is add a buffer in between them… but that makes the density file even bigger
[4:04 PM] beta_helix: Thank you all.. and keep up the great folding!